Article

# Multi-contrast large deformation diffeomorphic metric mapping for diffusion tensor imaging

Authors:
To read the full-text of this research, you can request a copy directly from the authors.

## Abstract

Diffusion tensor imaging (DTI) can reveal detailed white matter anatomy and has the potential to detect abnormalities in specific white matter structures. Such detection and quantification are, however, not straightforward. The voxel-based analysis after image normalization is one of the most widely used methods for quantitative image analyses. To apply this approach to DTI, it is important to examine if structures in the white matter are well registered among subjects, which would be highly dependent on employed algorithms for normalization. In this paper, we evaluate the accuracy of normalization of DTI data using a highly elastic transformation algorithm, called large deformation diffeomorphic metric mapping. After simulation-based validation of the algorithm, DTI data from normal subjects were used to measure the registration accuracy. To examine the impact of morphological abnormalities on the accuracy, the algorithm was also tested using data from Alzheimer's disease (AD) patients with severe brain atrophy. The accuracy level was measured by using manual landmark-based white matter matching and surface-based brain and ventricle matching as gold standard. To improve the accuracy level, cascading and multi-contrast approaches were developed. The accuracy level for the white matter was 1.88+/-0.55 and 2.19+/-0.84 mm for the measured locations in the controls and patients, respectively.

## No full-text available

... The reliability and accuracy of MRICloud for whole-brain segmentation, based on DTI or T1-WIs, have been extensively tested and validated (Ceritoglu et al., 2009;Liang et al., 2015;Oishi et al., 2008Oishi et al., , 2009Tang et al., 2015;Wu et al., 2016). A few other software that perform high-resolution T1-based automated segmentation, including FreeSurfer (Fischl, 2012), FSL (Jenkinson, Beckmann, Behrens, Woolrich, & Smith, 2012), SPM (Penny, Friston, Ashburner, Kiebel, & Nichols, 2007), ANTS (Avants et al., 2011), also underwent detailed reliability analysis, including testing the robustness of the respective pipelines to technical factors and artifacts (Ceritoglu et al., 2009;Han et al., 2006;Jovicich et al., 2009;Tustison et al., 2014;Ye et al., 2018). ...
... The reliability and accuracy of MRICloud for whole-brain segmentation, based on DTI or T1-WIs, have been extensively tested and validated (Ceritoglu et al., 2009;Liang et al., 2015;Oishi et al., 2008Oishi et al., , 2009Tang et al., 2015;Wu et al., 2016). A few other software that perform high-resolution T1-based automated segmentation, including FreeSurfer (Fischl, 2012), FSL (Jenkinson, Beckmann, Behrens, Woolrich, & Smith, 2012), SPM (Penny, Friston, Ashburner, Kiebel, & Nichols, 2007), ANTS (Avants et al., 2011), also underwent detailed reliability analysis, including testing the robustness of the respective pipelines to technical factors and artifacts (Ceritoglu et al., 2009;Han et al., 2006;Jovicich et al., 2009;Tustison et al., 2014;Ye et al., 2018). Most of these segmentation tools perform admirably when compared with the "gold standard" manual segmentation of selected structures, particularly when tested by the developers, in healthy subjects. ...
... The automated DTI segmentation was similar to that used for T1-WIs, except for the use of complementary contrasts (mean diffusivity [MD], fractional anisotropy [FA], and eigenvector [fiber orientation]) and a diffeomorphic likelihood fusion algorithm (Tang et al., 2014) for multi-atlas mapping. Please read (Ceritoglu et al., 2009) for technical details. We used the only atlas library available for DTI mapping in MRICloud: "Adults_168labels_12atlases_V1," which contains 12 healthy individuals, 20-50 years old. ...
Article
Full-text available
Introduction: The increasing use of large sample sizes for population and personalized medicine requires high-throughput tools for imaging processing that can handle large amounts of data with diverse image modalities, perform a biologically meaningful information reduction, and result in comprehensive quantification. Exploring the reproducibility of these tools reveals the specific strengths and weaknesses that heavily influence the interpretation of results, contributing to transparence in science. Methods: We tested-retested the reproducibility of MRICloud, a free automated method for whole-brain, multimodal MRI segmentation and quantification, on two public, independent datasets of healthy adults. Results: The reproducibility was extremely high for T1-volumetric analysis, high for diffusion tensor images (DTI) (however, regionally variable), and low for resting-state fMRI. Conclusion: In general, the reproducibility of the different modalities was slightly superior to that of widely used software. This analysis serves as a normative reference for planning samples and for the interpretation of structure-based MRI studies.
... Therefore, this imaging tool meets the requirements for a neuroimaging tool that is widely applicable to large-scale multimodal processing [13,14]. Furthermore, the reliability and accuracy of MRICloud for whole-brain segmentation, based T1-WIs, have been extensively validated [14][15][16][17][18][19][20]. Also, the test-retest reproducibility of MRICloud structural quantification have shown that the reproducibility for T1-volumetric analysis was significantly higher to that obtained using other well-established methods such as FreeSurfer and CONN-SPM, suggesting that it serves also a reliable tool for the interpretation of structure-based MRI studies, such as volumetric measurements [14]. ...
... Frontal lobe, temporal lobe, parietal lobe, striatum, (SFG), middle frontal gyrus (MFG), inferior frontal gyrus (IFG), superior parietal gyrus (SPG), amygdala, hippocampus, caudate, putamen, thalamus, corpus callosum, cerebellum, cingulum-cingulate gyrus, cingulum-hippocampus (CGH), middle fronto-orbital gyrus, lateral fronto-orbital gyrus (LFOG), superior temporal gyrus (STG), superior temporal white matter (STWM), middle temporal gyrus (MTG), middle temporal white matter, inferior temporal gyrus (ITG), inferior temporal white matter, temporal sulcus, nucleus accumbens. As defined in previous studies in detail [15][16][17][18][19][20], volumetric results were calculated and sent as separate files automatically by the MRICloud system. MRICloud provides a fully automated cloud service for brain parcellation of MPRAGE images based on MultipleAtlas Likelihood Fusion algorithm, JHU multi-atlas inventories with 286 defined structures, and an Ontology Level Control technology (https://mricloud. ...
Article
Full-text available
Background: Quantitative analysis of the high-resolution T1-weighted images provides useful markers to measure anatomical changes during brain degeneration related to major depressive disorder (MDD). However, there are controversial findings regarding these volume alterations in MDD indicating even to increased volumes in some specific regions in MDD patients. Methods: This study is a case-controlled study including 23 depression patients and 15 healthy subject person and 20-38 years of age, who have been treated at the Neurology and Psychiatry Department here. We compared specific anatomic regions between drug-free MDD patients and control group through MRI-Cloud, which is a novel brain imaging method that enables to analyze multiple brain regions simultaneously. Results: We have found that frontal, temporal, and parietal hemispheric volumes and middle frontal gyrus, inferior frontal gyrus, superior parietal gyrus, cingulum-hippocampus, lateral fronto-orbital gyrus, superior temporal gyrus, superior temporal white matter, middle temporal gyrus subanatomic regions were significantly reduced bilaterally in MDD patients compared to the control group, while striatum, amygdala, putamen, and nucleus accumbens bilaterally increased in MDD group compared to the control group suggesting that besides the heterogeneity among studies, also comorbid factors such as anxiety and different personal traits could be responsible for these discrepant results. Conclusion: Our study gives a strong message that depression is associated with altered structural brain volumes, especially, in drug-free and first-episode MDD patients who present with similar duration and severity of depression while the role of demographic and comorbid risk factors should not be neglected.
... MTR images were generated as MTR=(M 0 −M t )/M 0 . From the diffusion MRI data, diffusion tensors were calculated using the log-linear fitting method implemented in MRtrix (http://www.mrtrix.org) at each voxel, and maps of mean and radial diffusivities and FA were generated, The mouse brain images were spatial normalized to an ex vivo MRI template (Chuang et al., 2011) using the large deformation diffeomorphic metric mapping (LDDMM) method (Ceritoglu et al., 2009) implemented in the DiffeoMap software (https://www.mristudio.org). The template images had been normalized to the ARA using landmark-based image mapping and LDDMM. ...
... The mapping was then applied to the original atlas MRI data to generate an MRI template registered to the ARA space. Using dual-channel LDDMM (Ceritoglu et al., 2009) based on tissue contrasts in the average DWI and FA images and the MRI template, the 3D MRI data acquired in this study were accurately normalized to the ARA space. ...
Article
Full-text available
1 H MRI maps brain structure and function non-invasively through versatile contrasts that exploit inhomogeneity in tissue micro-environments. Inferring histopathological information from magnetic resonance imaging (MRI) findings, however, remains challenging due to absence of direct links between MRI signals and cellular structures. Here, we show that deep convolutional neural networks, developed using co-registered multi-contrast MRI and histological data of the mouse brain, can estimate histological staining intensity directly from MRI signals at each voxel. The results provide three-dimensional maps of axons and myelin with tissue contrasts that closely mimic target histology and enhanced sensitivity and specificity compared to conventional MRI markers. Furthermore , the relative contribution of each MRI contrast within the networks can be used to optimize multi-contrast MRI acquisition. We anticipate our method to be a starting point for translation of MRI results into easy-to-understand virtual histology for neurobiologists and provide resources for validating novel MRI techniques. Editor's evaluation This paper demonstrates how MRI can be used to mimic histological measures. This is something that the field of MRI has dubbed virtual histology (or MR-histology) for a while, but this paper is the first convincing demonstration that it can be achieved.
... DiffeoMap is used for image transformation and relies on large deformation diffeomorphic metric mapping (LDDMM). ROIEditor uses the results of DiffeoMap to implement image analysis according to a single atlas, the regional or voxel level (Ceritoglu et al., 2009;Oishi et al., 2009). Atlas-based normalization was performed on the images to create a template with structural parcellation (Fig. 1). ...
... It is appropriate for instance tasks like fiber tracking, color mapping, 3D visualization, and tensor calculation. It can be downloaded free from web site and most operations can be performed with only a few clicks by using this program (MRIStudio, 2007;Ceritoglu et al., 2009;Oishi et al., 2009). According to the measurement results obtained, a general increase in the intracerebral ventricle volumes of the patient group was observed. ...
Article
Knowing the volumetric changes in brain can allow for the estimation of the disease progression of various neurodegenerative disorders. Many studies have been shown that the volumetric changes in the some brain structures especially including the dopaminergic neurons, in patients with Parkinson’s disease (PD). The objective of this study was to compare intracerebral ventricles volume in patients with PD and healthy subjects to compare an automated atlas-based method (MRIStudio software) and a manual method (ImageJ). T1-weighted brain Magnetic Resonance Imaging (MRI) data of 21 patients with PD and 20 healthy individuals were used to calculate the intracerebral ventricle volumes. Measurement results obtained by ImageJ were considered as the gold standard. We found a significant increase in the left occipital part of the lateral ventricle volume in the patients with PD compared to the control subjects (p < 0.05). Also, no significant difference was found between the two methods of measurement (p > 0.05), meaning that a substantial agreement was found between the results obtained with the atlas-based analysis and manual method. The present study showed that MRIStudio can be performed easily and accurately on routine MRI scans for which the total intracerebral ventricles volume is to be estimated in PD. We suggest that, the attained volume values of intracerebral ventricles may provide a precious data for volumetric dependences of the anatomical structures in several clinical conditions.
... [13][14][15] LDDMM uses a cascade process to correct the large displacement between the 2 corresponding images. [17,18] TS of thicker slices is challenging because of misregistration and partial volume effects. In this study, to reduce the partial volume effects, an algorithm called adaptive voxel matching [19] was applied. ...
... LDDMM is designed to cope with a large amount of deformation while retaining the topology of the object. [17] However, the TS technique for thicker slices even with LDDMM did cause artifacts due to misregistration or partial volume effects. The AdVM algorithm was adopted to reduce the subtraction artifacts on the TS image with thick slices. ...
Article
Full-text available
To evaluate the improvement of radiologist performance in detecting bone metastases at follow up low-dose computed tomography (CT) by using a temporal subtraction (TS) technique based on an advanced nonrigid image registration algorithm.Twelve patients with bone metastases (males, 5; females, 7; mean age, 64.8 ± 7.6 years; range 51-81 years) and 12 control patients without bone metastases (males, 5; females, 7; mean age, 64.8 ± 7.6 years; 51-81 years) were included, who underwent initial and follow-up CT examinations between December 2005 and July 2016. Initial CT images were registered to follow-up CT images by the algorithm, and TS images were created. Three radiologists independently assessed the bone metastases with and without the TS images. The reader averaged jackknife alternative free-response receiver operating characteristics figure of merit was used to compare the diagnostic accuracy.The reader-averaged values of the jackknife alternative free-response receiver operating characteristics figures of merit (θ) significantly improved from 0.687 for the readout without TS and 0.803 for the readout with TS (P value = .031. F statistic = 5.24). The changes in the absolute value of CT attenuations in true-positive lesions were significantly larger than those in false-negative lesions (P < .001). Using TS, segment-based sensitivity, specificity, accuracy, positive predictive value, and negative predictive value of the readout with TS were 66.7%, 98.9%, 94.4%, 90.9%, and 94.8%, respectively.The TS images can significantly improve the radiologist's performance in the detection of bone metastases on low-dose and relatively thick-slice CT.
... The DTI images were automatically pre-processed and segmented using MRICloud (Mori et al., 2016). 2 The tensor reconstruction and quality control followed the pipeline of DTIStudio (Jiang et al., 2006). 3 MRICloud is a web-based platform that uses a fully automated multiatlas image parcellation algorithm that combines the image transformation algorithm, Large Deformation Diffeometric Mapping, based on complementary contrasts (e.g., MD, RD, FA, and fiber orientation; Ceritoglu et al., 2009), and a likelihood fusion algorithm for DTI multiatlas mapping and parcellation (Tang et al., 2014). This generated 168 regions of interest (ROIs), from which DTI scalar metrics (3 eigenvalues) were extracted. ...
Article
Full-text available
Objective This study examined the association of lifetime experiences, measured by a cognitive reserve (CR) composite score composed of years of education, literacy, and vocabulary measures, to level and rate of change in white matter microstructure, as assessed by diffusion tensor imaging (DTI) measures. We also examined whether the relationship between the proxy CR composite score and white matter microstructure was modified by participant age, APOE -ε4 genetic status, and level of vascular risk. Methods A sample of 192 non-demented ( n = 166 cognitively normal, n = 26 mild cognitive impairment) older adults [mean age = 70.17 (SD = 8.5) years] from the BIOCARD study underwent longitudinal DTI (mean follow-up = 2.5 years, max = 4.7 years). White matter microstructure was quantified by fractional anisotropy (FA) and radial diffusivity (RD) values in global white matter tracts and medial temporal lobe (MTL) white matter tracts. Results Using longitudinal linear mixed effect models, we found that FA decreased over time and RD increased over time in both the global and MTL DTI composites, but the rate of change in these DTI measures was not related to level of CR. However, there were significant interactions between the CR composite score and age for global RD in the full sample, and for global FA, global RD, and MTL RD among those with normal cognition. These interactions indicated that among participants with a lower baseline age, higher CR composite scores were associated with higher FA and lower RD values, while among participants with higher age at baseline, higher CR composite scores were associated with lower FA and higher RD values. Furthermore, these relationships were not modified by APOE -ε4 genotype or level of vascular risk. Conclusion The association between level of CR and DTI measures differs by age, suggesting a possible neuroprotective effect of CR among late middle-aged adults that shifts to a compensatory effect among older adults.
... One important thread of registration methods adopt diffeomorphic transformation that mathematically is a global one-to-one smooth and continuous mapping with invertible derivatives. Widely used heteromorphic parameterization methods include distance metric mapping [6,10], DARTEL [1] and diffeomorphic demons [34]. ...
Article
Full-text available
Deformable medical image registration plays a crucial role in theoretical research and clinical application. Traditional methods suffer from low registration accuracy and efficiency. Recent deep learning-based methods have made significant progresses, especially those weakly supervised by anatomical segmentations. However, the performance still needs further improvement, especially for images with large deformations. This work proposes a novel deformable image registration method based on an attention-guided fusion of multi-scale deformation fields. Specifically, we adopt a separately trained segmentation network to segment the regions of interest to remove the interference from the uninterested areas. Then, we construct a novel dense registration network to predict the deformation fields of multiple scales and combine them for final registration through an attention-weighted field fusion process. The proposed contour loss and image structural similarity index (SSIM) based loss further enhance the model training through regularization. Compared to the state-of-the-art methods on three benchmark datasets, our method has achieved significant performance improvement in terms of the average Dice similarity score (DSC), Hausdorff distance (HD), Average symmetric surface distance (ASSD), and Jacobian coefficient (JAC). For example, the improvements on the SHEN dataset are 0.014, 5.134, 0.559, and 359.936, respectively.
... The mapping was then applied to the original atlas MRI data to generate an MRI template registered to the ARA space. Using dual-channel LDDMM (35) based on tissue contrasts in the average DWI and FA images and the MRI template, the 3D MRI data acquired in this study were accurately normalized to the ARA space. ...
Preprint
Full-text available
H MRI maps brain anatomy and pathology non-invasively through contrasts generated by exploiting inhomogeneities in tissue micro-environments. Inferring histopathological information from MRI findings, however, remains challenging due to the absence of direct links between MRI signals and specific tissue compartments. Here, we show that convolutional neural networks, developed using co-registered multi-contrast MRI and histological data of the mouse brain, can generate virtual histology from MRI results. Our networks provide maps that mirror histological stains for axons and myelin with enhanced specificity compared to existing MRI markers. Furthermore, by introducing random perturbations to the inputs, the relative contribution of each MRI contrast within the networks can be estimated and guide the optimization of MRI acquisition. We anticipate our method to be a starting point for translation of MRI results into easy-to-understand virtual histology for neurobiologists and provide resources for developing novel MRI contrasts.
... Following motion correction (28), the tensor field for each individual brain was calculated using DTIStudio (www.MriStudio.org) and automatically fit to JHU-MNI atlas space using Large Deformation 6 Diffeomorphic Metric Mapping (28,29). Fractional anisotropy (FA), and axial (AD, first eigenvalue) and radial diffusivity (RD, mean of second and third eigenvalues), were measured in anatomical regions defined in the JHU-MNI atlas (30). ...
Preprint
Full-text available
Objective Cognitive deficits and microstructural brain abnormalities are well documented in HIV-positive individuals (HIV+). This study evaluated whether chronic marijuana (MJ) use contributes to additional cognitive deficits or brain microstructural abnormalities that may reflect neuroinflammation or neuronal injury in HIV+. Method Using a 2 × 2 design, 44 HIV + participants, [23 minimal/no MJ users (HIV+), 21 chronic active MJ users (HIV + MJ)], were compared to 46 seronegative participants [24 minimal/no MJ users (SN) and 22 chronic MJ users (SN + MJ)] on neuropsychological performance (7 cognitive domains) and diffusion tensor imaging metrics, using an automated atlas to assess fractional anisotropy (FA), axial (AD), radial (RD), and mean (MD) diffusivities, in 18 cortical and 4 subcortical brain regions, Results Compared to SN and regardless of MJ use, the HIV + group had lower FA and higher diffusivities in multiple white matter and subcortical structures (p = 0.001–0.050), as well as poorer cognition in Fluency (p = 0.039), Attention / Working Memory (p = 0.009), Learning (p = 0.015) and Memory (p = 0.028). Regardless of HIV-serostatus, MJ users had lower AD in uncinate fasciculus (p = 0.016) but similar cognition as non-users. No additive or interactive effects were found between HIV-serostatus and MJ use on DTI metrics or cognitive function. Furthermore, higher MD in thalamus predicted poorer fluency, learning and memory in HIV + participants, while higher RD in posterior corona radiata predicted poorer learning in MJ users. Lower FA in the anterior internal capsule also predicted worse attention/working memory in all except SN subjects. Lastly, MJ users with or without HIV-infection showed greater than normal age-dependent FA declines in superior longitudinal fasciculus, external capsule and globus pallidus.
... Three eigenvalues and eigenvectors were obtained from the tensor field, from which the fractional anisotropy (FA) and mean diffusivity (MD) maps were generated. DTI maps in the original space were transformed to the JHU-MNI atlas space using a 12-parameter linear transformation followed by dual-channel large deformation diffeomorphic metric mapping (LDDMM) (Ceritoglu et al. 2009). The inverse transformations generated with this process were applied to the JHU-MNI atlas Type II parcellation map to obtain 130 regions in native space, from which the FA and MD of each region was calculated ). ...
Article
Full-text available
AimsCognitive impairment may be greater in HIV-positive (HIV+) women than in HIV+ men. Whether sex-specific differences exist in brain microstructure of HIV+ individuals is unknown and was evaluated.Method39 HIV+ (21 men, 18 women) and 45 seronegative (SN, 20 men, 25 women) participants were assessed with brain diffusion tensor imaging and cognitive assessments (7 neuropsychological domains). Fractional anisotropy (FA) and mean diffusivity (MD) were measured with an automated atlas in selected brain regions. Group comparisons were assessed with linear mixed effects models, with sub-regions and hemisphere (left/right) as repeated factors for each region.ResultsHIV+ women, but not HIV+ men, were slower than sex-matched SN controls on sensorimotor function (Dominant-hand: interaction-p = 0.007; Non-dominant hand: interaction-p = 0.039). Similarly, only HIV+ women had lower FA in the globus pallidus (GP, interaction-p = 0.011). Additionally, regardless of sex, the HIV+ group had poorer Fluency, Speed, and Attention than SN-controls (p = 0.006–0.008), as well as lower FA and higher MD in multiple brain regions (p = <0.001–0.044). Across all participants, performance on Attention was predicted by uncinate-FA (p < 0.001, r = 0.5) and corpus callosum (CC)-FA (p = 0.038, r = 0.23), while the Speed of Information Processing was predicted by CC-FA (p = 0.009, r = 0.3). Furthermore, faster sensorimotor function correlated with higher CC-FA and uncinate-FA in men but not in women (Sex*DTI-interaction-p = 0.03–0.06).Conclusions The relatively poorer sensorimotor function and abnormally lower GP_FA, suggesting lesser neuronal integrity, in HIV+ women demonstrate sex-specific effects from HIV-infection on these measures. These findings may be related to the greater immune activation and neuroinflammation in HIV+ women compared to HIV+ men. Graphical Abstract
... All MRIs were bias-corrected and linearly aligned to the Johns Hopkins University JHU-MNI atlas space (Oishi, 2009). Then, atlases in the JHU T1 Geriatric Multi-Atlas Inventory (Djamanakova, 2013;Wu, 2016), which is designed for geriatric patient populations with potential brain atrophy, were transformed to the linearly aligned subject image using large deformation diffeomorphic metric mapping (LDDMM) (Ceritoglu, 2009). The multi-atlas fusion algorithm (Tang, 2013) was applied to the transformed atlases to obtain a brain-structure parcellation map specific to each subject MRI. ...
Article
Background To examine the interaction between structural brain volume measures derived from a clinical magnetic resonance imaging (MRI) and occurrence of neuropsychiatric symptoms (NPS) in outpatient memory clinic patients. Methods Clinical and neuroimaging data were collected from the medical records of outpatient memory clinic patients who were seen by neurologists, geriatric neuropsychiatrists, and geriatricians. MRI scan acquisition was carried out on a 3 T Siemens Verio scanner at Johns Hopkins Bayview Medical Center. Image analyses used an automated multi-label atlas fusion method with a geriatric atlas inventory to generate 193 anatomical regions from which volumes were measured. Regions of interest were generated a priori based on previous literature review of NPS in dementia. Regional volumes for agitation, apathy, and delusions were carried forward in a linear regression analysis. Results Seventy-two patients had clinical and usable neuroimaging data that were analyzed and grouped by Mini-Mental State Exam (MMSE). Neuropsychiatric Inventory Questionnaire (NPI-Q) agitation was inversely associated with rostral anterior cingulate cortex (ACC) bilaterally and left subcallosal ACC volumes in the moderate severity group. Delusions were positively associated with left ACC volumes in both severe and mild groups but inversely associated with the right dorsolateral prefrontal cortex (DLPFC) in the moderate subgroup. Conclusions Agitation, apathy, and delusions are associated with volumes of a priori selected brain regions using clinical data and clinically acquired MRI scans. The ACC is an anatomic region common to these symptoms, particularly agitation and delusions, which closely mirror the findings of research-quality studies and suggest its importance as a behavioral hub.
... Several studies propose pairwise optimization methods for non-rigid image registration within displacement vector fields, including elastic type models, free-form deformations with b-splines [40], discrete methods [11,16] and Demons [33,42]. There are also several methods proposing diffeomorphic transformation-based methods including Large Diffeomorphic Distance Metric Mapping (LDDMM) [4,45,8,9,17,20,31], DARTEL [1] and diffeomorphic demons [43]. These methods are not learning based and need to be repeated for each pair, which is time-consuming when dealing with large data set. ...
Preprint
Image registration algorithms can be generally categorized into two groups: non-rigid and rigid. Recently, many deep learning-based algorithms employ a neural net to characterize non-rigid image registration function. However, do they always perform better? In this study, we compare the state-of-art deep learning-based non-rigid registration approach with rigid registration approach. The data is generated from Kaggle Dog vs Cat Competition \url{https://www.kaggle.com/c/dogs-vs-cats/} and we test the algorithms' performance on rigid transformation including translation, rotation, scaling, shearing and pixelwise non-rigid transformation. The Voxelmorph is trained on rigidset and nonrigidset separately for comparison and we also add a gaussian blur layer to its original architecture to improve registration performance. The best quantitative results in both root-mean-square error (RMSE) and mean absolute error (MAE) metrics for rigid registration are produced by SimpleElastix and non-rigid registration by Voxelmorph. We select representative samples for visual assessment.
... DTI reconstruction and quality control were also performed using MRICloud, which follows the pipeline of DTIStudio (Jiang et al., 2006). MRICloud offers a fully automated multiatlas image parcellation algorithm, which combines the image transformation algorithm, Large Deformation Diffeomorphic Metric Mapping Grenander and Miller, 1998;Miller et al., 1997) based on complementary contrasts (mean diffusivity [MD], fractional anisotropy [FA], and fiber orientation; Ceritoglu et al., 2009), and a likelihood fusion algorithm for DTI multiatlas mapping and parcellation . The DTI multiatlas template set contains 12 healthy adult brains and results in the parcellation of 168 brain structures, from which vectors of DTI scalar metrics (3 eigenvalues) and volumes were extracted. ...
Article
Significant evidence demonstrates that aging is associated with variability in cognitive performance, even among individuals who are cognitively normal. In this study, we examined measures from magnetic resonance imaging and cerebrospinal fluid (CSF) to investigate which measures, alone or in combination, were associated with individual differences in episodic memory performance. Using hierarchical linear regressions, we compared the ability of diffusion tensor imaging (DTI) metrics, CSF measures of amyloid and tau, and gray matter volumes to explain variability in memory performance in a cohort of cognitively normal older adults. Measures of DTI microstructure were significantly associated with variance in memory performance, even after accounting for the contribution of the CSF and magnetic resonance imaging gray matter volume measures. Significant associations were found between DTI measures of the hippocampal cingulum and fornix with individual differences in memory. No such relationships were found between memory performance and CSF markers or gray matter volumes. These findings suggest that DTI metrics may be useful in identifying changes associated with aging or age-related diseases.
... DKI data were obtained using a single-shot twice-refocused 2D spin-echo echo-planar imaging sequence with TE/TR: 98/8600, contiguous slices with 3 mm thickness, 3 mm isotropic resolution, one signal average, 30 noncollinear diffusion-weighted gradient directions, 3 diffusion weightings (b = 0, 1000, 2000 s/mm 2 ), generalized autocalibrating partially parallel acquisition (GRAPPA) acceleration factor of 2.0, and an acquisition time of ~9.1 minutes. Both diffusion-weighted (b = 1000, 2000 s/mm 2 ) and non-diffusion-weighted (b = 0 s/mm 2 ) data were processed using diffusion kurtosis estimator (DKE) [19] and DiffeoMap [20] and analyzed using ROIEditor [21] [ Figure 1]. Briefly, first, after processing with DKE, we obtained four DKI-based metrics, that is, AK, RK, MK, and KFA. ...
Article
Full-text available
Purpose: The association between obstructive sleep apnea (OSA) and cognitive impairment is well-recognized, but little is known about neural derangements that underlie this phenomenon. The purpose of this study was to evaluate the utility of diffusion kurtosis imaging (DKI) using a whole-brain atlas to comprehensively assess microstructural tissue changes in the brain of patients with OSA. Methods: This prospective study was conducted in 20 patients with moderate-to-severe OSA and 20 age- and gender-matched controls. MRI data acquisition was performed with 3 Tesla and data was analyzed using a whole-brain atlas. DKI data were processed and transformed into a brain template space to obtain various kurtosis parameters including axial kurtosis (AK), radial kurtosis (RK), mean kurtosis (MK), and kurtosis fractional anisotropy (KFA) using a 189-region brain atlas in the same template space. These kurtosis measurements were further analyzed using a student t-test in order to determine kurtosis measurements that present significant differences between the OSA patient set and the control set. Results: Significant differences (P < 0.05) were found in AK (54 regions), RK (10 regions), MK (6 regions) and KFA (41 regions) values in patients with OSA as compared to controls. DKI indices, using an atlas-based whole-brain analysis approach used in our study, showed widespread involvement of the anatomical regions in patients with OSA. Conclusion: The kurtosis parameters are more sensitive in demonstrating abnormalities in brain tissue structural organization at the microstructural level before any detectable changes appear in conventional MRI or other imaging modalities.
... We studied the effect of registration methods as the accuracy of image registration significantly influences the quantitative analysis results. Our experiments showed that multicontrast registration (FA and b0) was substantially more accurate than single contrast (FA only) registration, in particular outside of the white matter, consistently with the conclusions made by (Ceritoglu et al. 2009). This could be expected because FA does not provide much information outside the WM. ...
Article
Full-text available
Diffusion MRI is the modality of choice to study alterations of white matter. In past years, various works have used diffusion MRI for automatic classification of Alzheimer’s disease. However, classification performance obtained with different approaches is difficult to compare because of variations in components such as input data, participant selection, image preprocessing, feature extraction, feature rescaling (FR), feature selection (FS) and cross-validation (CV) procedures. Moreover, these studies are also difficult to reproduce because these different components are not readily available. In a previous work (Samper-González et al. 2018), we propose an open-source framework for the reproducible evaluation of AD classification from T1-weighted (T1w) MRI and PET data. In the present paper, we first extend this framework to diffusion MRI data. Specifically, we add: conversion of diffusion MRI ADNI data into the BIDS standard and pipelines for diffusion MRI preprocessing and feature extraction. We then apply the framework to compare different components. First, FS has a positive impact on classification results: highest balanced accuracy (BA) improved from 0.76 to 0.82 for task CN vs AD. Secondly, voxel-wise features generally gives better performance than regional features. Fractional anisotropy (FA) and mean diffusivity (MD) provided comparable results for voxel-wise features. Moreover, we observe that the poor performance obtained in tasks involving MCI were potentially caused by the small data samples, rather than by the data imbalance. Furthermore, no extensive classification difference exists for different degree of smoothing and registration methods. Besides, we demonstrate that using non-nested validation of FS leads to unreliable and over-optimistic results: 5% up to 40% relative increase in BA. Lastly, with proper FR and FS, the performance of diffusion MRI features is comparable to that of T1w MRI. All the code of the framework and the experiments are publicly available: general-purpose tools have been integrated into the Clinica software package (www.clinica.run) and the paper-specific code is available at: https://github.com/aramis-lab/AD-ML.
... There is extensive work in deformable image registration methods [5,6,7,10,19,28,67,71,73]. Conventional frameworks optimize a regularized dense deformation field that matches one image with the other [7,67]. Diffeomorphic transforms are toplogy preserving and invertible, and have been widely used in computational neuroanatomy analysis [6,5,10,13,14,32,41,55,59,69,73]. While extensively studied, conventional registration algorithms require an optimization for every pair of images, leading to long runtimes in practice. ...
Preprint
We develop a learning framework for building deformable templates, which play a fundamental role in many image analysis and computational anatomy tasks. Conventional methods for template creation and image alignment to the template have undergone decades of rich technical development. In these frameworks, templates are constructed using an iterative process of template estimation and alignment, which is often computationally very expensive. Due in part to this shortcoming, most methods compute a single template for the entire population of images, or a few templates for specific sub-groups of the data. In this work, we present a probabilistic model and efficient learning strategy that yields either universal or conditional templates, jointly with a neural network that provides efficient alignment of the images to these templates. We demonstrate the usefulness of this method on a variety of domains, with a special focus on neuroimaging. This is particularly useful for clinical applications where a pre-existing template does not exist, or creating a new one with traditional methods can be prohibitively expensive. Our code and atlases are available online as part of the VoxelMorph library at http://voxelmorph.csail.mit.edu.
... Samples were prepared according to the method described by Tyszka et al. 24 Brains were imaged using 20 distinct gradient directions on a 9.4T Bruker Biospec MRI scanner (Bruker BioSpin). All DTI processing and connectome analyses were performed as described by Sahnoune et al. 25 Briefly, ROIEditor, DTI studio, 26,27 template maps from Duke Center for In vivo Microscopy, 28 DiffeoMap, 29 AIR algorithm, 30,31 and large deformation diffeomorphic metric mapping (LDDMM) 32 were used for alignment, segmentation, and calculation. DTI studio software (www.mristudio.org) was used to generate various volume maps and calculate the average FA value. ...
Article
Full-text available
Background Cranial radiotherapy (CRT) is an important part of brain tumor treatment, and while highly effective, survivors suffer from long-term cognitive side-effects. In this study we aim to establish late-term imaging-markers of CRT-induced brain injury and identify functional markers indicative of cognitive performance. Specifically, we aim to identify changes in executive function, brain metabolism, and neuronal organization. Methods Male Sprague Dawley rats were fractionally irradiated at 28 days of age to a total dose of 30 Gy to establish a radiation-induced brain injury model. Animals were trained at 3 months post-CRT using the 5-choice serial reaction time task. At 12 months post-CRT animals were evaluated for cognitive and imaging changes which included PET/CT and MRI. Results Cognitive deficit with signs of neuroinflammation were found at 12 months post-CRT in irradiated animals. CRT resulted in significant volumetric changes in 38% of brain regions as well as overall decrease in brain volume and reduced gray matter volume. PET imaging showed higher brain glucose uptake in CRT animals. Using MRI, irradiated brains had an overall decrease in fractional anisotropy, lower global efficiency, increased transitivity, and altered regional connectivity. Cognitive measurements were found to be significantly correlated with six image features which included myelin integrity and local organization of the neural network. Conclusions These results demonstrate that CRT leads to late-term morphological changes, reorganization of neural connections, and metabolic dysfunction. The correlation between imaging-markers and cognitive deficits can be used to assess late-term side-effects of brain tumor treatment and evaluate efficacy of new interventions.
... For each developmental stage, group average images were generated from the 3D data using the iterative methods described in (Chuang et al. 2011;Kovacevic et al. 2005), first using intensity-based linear affine transformation and then dual channel (aDW + FA) Large Deformation Diffeomorphic Metric Mapping (LDDMM) (Ceritoglu et al. 2009) implemented in Diffeomap (http://www.mrist udio.org), which took approximately 24 h on a linux cluster. ...
Article
Full-text available
Diffusion tensor imaging (DTI) is increasingly utilized as a sensitive tool for studying brain maturation and injuries during the neonatal period. In this study, we acquired high resolution in vivo DTI data from neonatal rat brains from postnatal day 2 (P2) to P10 and correlated temporal changes in DTI derived markers with microstructural organization of glia, axons, and dendrites during this critical period of brain development. Group average images showed dramatic temporal changes in brain morphology, fractional anisotropy (FA) and mean diffusivity (MD). Most cortical regions showed a monotonous decline in FA and an initial increase in MD from P2 to P8 that declined slightly by P10. Qualitative histology revealed rapid maturation of the glial and dendritic networks in the developing cortex. In the cingulate and motor cortex, the decreases in FA over time significantly correlated with structural anisotropy values computed from histological sections stained with glial and dendritic markers. However, in the sensory and visual cortex, other factors probably contributed to the observed decreases in FA. We did not observe any significant correlations between FA and structural anisotropy computed from the axonal histological marker.
... Then, skull stripping of all the images was performed by generating participant-specific brain masks in SPM12, and refining the masks manually using the ROIEditor Toolbox in MRIStudio. The coregistered and skull-stripped mean images were then normalized to the "JHU_MNI_SS_T1_ss" template 51 in Montreal Neurological Institute (MNI) coordi- nate space 52 using a 12-parameter affine (linear) transformation with Automated Image Registration (AIR), followed by high-dimensional, non-linear warping with the large deformation diffeomorphic metric mapping (LDDMM) algorithm with alpha values of 0.01, 0.005, and 0.002 53 in MRIStudio's DiffeoMap Toolbox, as previ- ously reported 54 . The alpha values constrain the amount of elasticity allowed in each iteration of the deformation, so using three iterations with cascading alpha values allows for increasingly non-linear registrations. ...
... Then, skull stripping of all the images was performed by generating participant-specific brain masks in SPM12, and refining the masks manually using the ROIEditor Toolbox in MRIStudio. The coregistered and skull-stripped mean images were then normalized to the "JHU_MNI_SS_T1_ss" template 51 in Montreal Neurological Institute (MNI) coordinate space 52 using a 12-parameter affine (linear) transformation with Automated Image Registration (AIR), followed by high-dimensional, non-linear warping with the large deformation diffeomorphic metric mapping (LDDMM) algorithm with alpha values of 0.01, 0.005, and 0.002 53 in MRIStudio's DiffeoMap Toolbox, as previously reported 54 . The alpha values constrain the amount of elasticity allowed in each iteration of the deformation, so using three iterations with cascading alpha values allows for increasingly non-linear registrations. ...
Article
Full-text available
Various MRI techniques, including myelin water imaging, T1w/T2w ratio mapping and diffusion-based imaging can be used to characterize tissue microstructure. However, surprisingly few studies have examined the degree to which these MRI measures are related within and between various brain regions. Therefore, whole-brain MRI scans were acquired from 31 neurologically-healthy participants to empirically measure and compare myelin water fraction (MWF), T1w/T2w ratio, fractional anisotropy (FA), axial diffusivity (AD), radial diffusivity (RD) and mean diffusivity (MD) in 25 bilateral (10 grey matter; 15 white matter) regions-of-interest (ROIs). Except for RD vs. T1w/T2w, MD vs. T1w/T2w, moderately significant to highly significant correlations (p < 0.001) were found between each of the other measures across all 25 brain structures [T1w/T2w vs. MWF (Pearson r = 0.33, Spearman ρ = 0.31), FA vs. MWF (r = 0.73, ρ = 0.75), FA vs. T1w/T2w (r = 0.25, ρ = 0.22), MD vs. AD (r = 0.57, ρ = 0.58), MD vs. RD (r = 0.64, ρ = 0.61), AD vs. MWF (r = 0.43, ρ = 0.36), RD vs. MWF (r = −0.49, ρ = −0.62), MD vs. MWF (r = −0.22, ρ = −0.29), RD vs. FA (r = −0.62, ρ = −0.75) and MD vs. FA (r = −0.22, ρ = −0.18)]. However, while all six MRI measures were correlated with each other across all structures, there were large intra-ROI and inter-ROI differences (i.e., with no one measure consistently producing the highest or lowest values). This suggests that each quantitative MRI measure provides unique, and potentially complimentary, information about underlying brain tissues – with each metric offering unique sensitivity/specificity tradeoffs to different microstructural properties (e.g., myelin content, tissue density, etc.).
... They produced DtiStudio in 2001 [37,38]. Users can now obtain a singlesubject atlas containing at least 286 brain regions to produce automated brain segmentation [38,39]. ...
Chapter
Full-text available
Since the development of the neuroimaging sciences, the posterior cranial fossa (PCF) has been anatomically evaluated and measured volumetrically. Radiologically, the PCF volume is an important neuroimaging finding in Chiari malformation type I (CM I), but recent neuroimaging studies have revealed conflicting volumetric changes in the PCF in CM I patients. In this chapter, anatomical details of the PCF, different automated imaging techniques, and volumetric changes in the PCF will be reviewed. The main purpose is to gather information about automated imaging methods for the PCF and techniques for measuring brain substructures in patients with CM I and their clinical significance.
... DTIStudio (www.mristudio.org) was used to align all DWIs to the average of b 0 s to remove small sample displacements due to vibrations during the long scan and compute average DWI (aDWI) and fractional anisotropy (FA) maps using the diffusion tensor model. The aDWI and FA maps were normalized to an MRI-based atlas (Chuang et al., 2011, Arefin et al., 2019 using the dual-channel (aDWI+FA) large deformation diffeomorphic metric mapping (LDDMM) (Ceritoglu et al., 2009). To assess global and regional amygdala connectivity, we transferred 14 structural labels (i.e. 7 structures in each hemisphere, e.g., cortex, hippocampus, cerebellum, etc. as defined in the Allen mouse brain atlas) to the subject images using the inverse mapping from LDDMM. ...
... This method uses Gaussian smoothing as a regularization of the displacement field and additive accumulation during the iterative process. What is more, diffeomorphic transforms [33], which preserve topology and invertibility on the transformation, have shown remarkable superiority in various computational anatomy studies [34][35][36][37][38]. Actually, there have been many learning-based registration methods, such as Quicksilver [39], VoxelMorph [40], and BIRNet [41]. ...
Article
Full-text available
Multimodal registration is a challenging task due to the significant variations exhibited from images of different modalities. CT and MRI are two of the most commonly used medical images in clinical diagnosis, since MRI with multicontrast images, together with CT, can provide complementary auxiliary information. The deformable image registration between MRI and CT is essential to analyze the relationships among different modality images. Here, we proposed an indirect multimodal image registration method, i.e., sCT-guided multimodal image registration and problematic image completion method. In addition, we also designed a deep learning-based generative network, Conditional Auto-Encoder Generative Adversarial Network, called CAE-GAN, combining the idea of VAE and GAN under a conditional process to tackle the problem of synthetic CT (sCT) synthesis. Our main contributions in this work can be summarized into three aspects: (1) We designed a new generative network called CAE-GAN, which incorporates the advantages of two popular image synthesis methods, i.e., VAE and GAN, and produced high-quality synthetic images with limited training data. (2) We utilized the sCT generated from multicontrast MRI as an intermediary to transform multimodal MRI-CT registration into monomodal sCT-CT registration, which greatly reduces the registration difficulty. (3) Using normal CT as guidance and reference, we repaired the abnormal MRI while registering the MRI to the normal CT.
... Therefore, this imaging tool meets the requirements for a neuroimaging tool that is widely applicable to large-scale multimodal processing [22,23]. Furthermore, the reliability and accuracy of MRI-Cloud for whole-brain segmentation, based T1-WIs, have been extensively validated [23][24][25][26][27][28][29]. Also, the test-retest reproducibility of MRI-Cloud structural quantification has shown that the reproducibility for T1-volumetric analysis was significantly higher than that obtained using other well-established methods such as Free Surfer and CONN-SPM, suggesting that it serves also a reliable tool for the interpretation of structure-based MRI studies, such volumetric measurements [23]. ...
... Following motion correction [28], the tensor field for each individual brain was calculated using DTIStudio (www.MriStudio.org) and automatically fit to JHU-MNI atlas space using Large Deformation Diffeomorphic Metric Mapping [28,29]. Fractional anisotropy (FA), and axial (AD, first eigenvalue), and radial diffusivity (RD, mean of second and third eigenvalues) were measured in anatomical regions defined in the JHU-MNI atlas [30]. ...
Article
Full-text available
Objective: Cognitive deficits and microstructural brain abnormalities are well documented in HIV-positive individuals (HIV+). This study evaluated whether chronic marijuana (MJ) use contributes to additional cognitive deficits or brain microstructural abnormalities that may reflect neuroinflammation or neuronal injury in HIV+. Method: Using a 2 × 2 design, 44 HIV+ participants [23 minimal/no MJ users (HIV+), 21 chronic active MJ users (HIV + MJ)] were compared to 46 seronegative participants [24 minimal/no MJ users (SN) and 22 chronic MJ users (SN + MJ)] on neuropsychological performance (7 cognitive domains) and diffusion tensor imaging metrics, using an automated atlas to assess fractional anisotropy (FA), axial (AD), radial (RD), and mean (MD) diffusivities, in 18 cortical and 4 subcortical brain regions. Results: Compared to SN and regardless of MJ use, the HIV+ group had lower FA and higher diffusivities in multiple white matter and subcortical structures (p < 0.001-0.050), as well as poorer cognition in Fluency (p = 0.039), Attention/Working Memory (p = 0.009), Learning (p = 0.014), and Memory (p = 0.028). Regardless of HIV serostatus, MJ users had lower AD in uncinate fasciculus (p = 0.024) but similar cognition as nonusers. HIV serostatus and MJ use showed an interactive effect on mean diffusivity in the right globus pallidus but not on cognitive function. Furthermore, lower FA in left anterior internal capsule predicted poorer Fluency across all participants and worse Attention/Working Memory in all except SN subjects, while higher diffusivities in several white matter tracts also predicted lower cognitive domain Z-scores. Lastly, MJ users with or without HIV infection showed greater than normal age-dependent FA declines in superior longitudinal fasciculus, external capsule, and globus pallidus. Conclusions: Our findings suggest that, except in the globus pallidus, chronic MJ use had no additional negative influence on brain microstructure or neurocognitive deficits in HIV+ individuals. However, lower AD in the uncinate fasciculus of MJ users suggests axonal loss in this white matter tract that connects to cannabinoid receptor rich brain regions that are involved in verbal memory and emotion. Furthermore, the greater than normal age-dependent FA declines in the white matter tracts and globus pallidus in MJ users suggest that older chronic MJ users may eventually have lesser neuronal integrity in these brain regions.
... A diffeomorphic deformation ensures certain desirable properties between two image volumes like continuous, differentiable, and preserving topology [6], [7], [21]. The popular examples of these algorithmic extensions to large deformation are Large Deformation Diffeomorphic Metric Matching (LDDMM) [6], [22]- [24] and Symmetric image Normalization method (SyN) [7]. ...
Preprint
Image registration is a fundamental task in medical image analysis. Recently, deep learning based image registration methods have been extensively investigated due to their excellent performance despite the ultra-fast computational time. However, the existing deep learning methods still have limitation in the preservation of original topology during the deformation with registration vector fields. To address this issues, here we present a cycle-consistent deformable image registration. The cycle consistency enhances image registration performance by providing an implicit regularization to preserve topology during the deformation. The proposed method is so flexible that can be applied for both 2D and 3D registration problems for various applications, and can be easily extended to multi-scale implementation to deal with the memory issues in large volume registration. Experimental results on various datasets from medical and non-medical applications demonstrate that the proposed method provides effective and accurate registration on diverse image pairs within a few seconds. Qualitative and quantitative evaluations on deformation fields also verify the effectiveness of the cycle consistency of the proposed method.
... It is essential to choose a suitable differential symmetric registration algorithm [12]. Commonly used heteromorphic parameterization methods include distance metric mapping (LDDMM) [13][14][15][16][17], DARTEL [18], diffeomorphic demons [19] and standard symmetric normalization (SyN) [20]. The main problem faced by traditional registration methods is: for each image to be registered, iteratively optimizes the cost function from scratch, which severely limits registration speed and ignores the inherent registration mode shared between the same data [21]. ...
Article
Full-text available
Traditional deformable registration methods achieve brilliant results and show strong theoretical support but are computational intensive since they optimize each image pair’s objective function. Recently, supervised learning methods have facilitated fast registration. However, it requires ground truth and does not guarantee a diffeomorphism registration. This paper proposes a new unsupervised learning method Recursive Cascaded Network with a segmentation mask, for two-dimensional medical image registration. Different from the original cascaded network, the network framework into two parts. The first section obtains a pair of image rolls and uses the registration sub-network to predict the deformation vector field from the moving image to the fixed image. The second part introduces anatomical segmentation into the network during training, makes full use of the auxiliary information of the volume, adds an autoencoder to encode the anatomical segmentation, and incorporates it into the learning process of the model in the form of constraints. The local and global ideas are combined to ensure the deformation field’s rationality and improve the distribution. The most important thing is that we propose a formula for calculating the cascaded network’s deformation field used in the test stage to evaluate the relationship between the registration accuracy and the deformation field’s effectiveness. Our experiments show that the system has a better registration effect and less information loss than the current state-of-the-art method. Simultaneously, the cascade method’s accuracy is an improvement at a certain number of layers, and the increase in accuracy needs to sacrifice the effectiveness of the deformation field.
... In order to quantify the embryonic brain deformation with development, we performed longitudinal registration between consecutive embryonic stages. For instance, the E11.5 atlas was transformed to the E12.5 atlas, first by landmark-based rigid transformation and intensity-based rigid transformation and then by the two-channel LDDMM based on the mDWI and FA contrasts (79). In the cost function used by LDDMM, the term that controls the smoothness of the vector field was adjusted to capture macroscopic changes (>0.1 mm) in brain morphology. ...
Article
Full-text available
The embryonic mouse brain undergoes drastic changes in establishing basic anatomical compartments and laying out major axonal connections of the developing brain. Correlating anatomical changes with gene-expression patterns is an essential step toward understanding the mechanisms regulating brain development. Traditionally, this is done in a cross-sectional manner, but the dynamic nature of development calls for probing gene–neuroanatomy interactions in a combined spatiotemporal domain. Here, we present a four-dimensional (4D) spatiotemporal continuum of the embryonic mouse brain from E10.5 to E15.5 reconstructed from diffusion magnetic resonance microscopy (dMRM) data. This study achieved unprecedented high-definition dMRM at 30- to 35-µm isotropic resolution, and together with computational neuroanatomy techniques, we revealed both morphological and microscopic changes in the developing brain. We transformed selected gene-expression data to this continuum and correlated them with the dMRM-based neuroanatomical changes in embryonic brains. Within the continuum, we identified distinct developmental modes comprising regional clusters that shared developmental trajectories and similar gene-expression profiles. Our results demonstrate how this 4D continuum can be used to examine spatiotemporal gene–neuroanatomical interactions by connecting upstream genetic events with anatomical changes that emerge later in development. This approach would be useful for large-scale analysis of the cooperative roles of key genes in shaping the developing brain.
... DTIStudio (http://www.mristudio.org) was used to align all DWIs to the average of b 0 s to remove small sample displacements due to vibrations during the long scan and compute average DWI (aDWI) and FA maps using the diffusion tensor model. The aDWI and FA maps were normalized to an MRI-based atlas (Chuang et al., 2011;Arefin et al., 2019) using the dual-channel (aDWI+FA) large deformation diffeomorphic metric mapping (LDDMM) (Ceritoglu et al., 2009). To assess global and regional AMY connectivity, we transferred 28 structural labels (i.e. 14 nodes and 182 connectomes in each hemisphere) to the subject images using the inverse mapping from LDDMM. ...
Article
Full-text available
It is currently unclear whether early life stress (ELS) affects males and females differently. However, a growing body of work has shown that sex moderates responses to stress and injury, with important insights into sex-specific mechanisms provided by work in rodents. Unfortunately, most of the ELS studies in rodents were conducted only in males, a bias that is particularly notable in translational work that has used human imaging. Here we examine the effects of unpredictable postnatal stress (UPS), a mouse model of complex ELS, using high resolution diffusion magnetic resonance imaging. We show that UPS induces several neuroanatomical alterations that were seen in both sexes and resemble those reported in humans. In contrast, exposure to UPS induced fronto-limbic hyper-connectivity in males, but either no change or hypoconnectivity in females. Moderated-mediation analysis found that these sex-specific changes are likely to alter contextual freezing behavior in males but not in females.
... Distortion corrected and anonymized MRIs were uploaded to MRICloud [https://mricloud.org; Mori et al., 2016] for segmentation using the fully automated MRICloud T1-Segmentation pipeline (version 7A) [Ceritoglu et al., 2009;Wang & Yushkevich, 2013]. The whole-brain segmentation output for each participant was visually inspected slice-wise for segmentation quality. ...
Article
Atypical responses to fearful stimuli and the presence of various forms of anxiety are commonly seen in children with autism spectrum disorder (ASD). The fear potentiated startle paradigm (FPS), which has been studied both in relation to anxiety and as a probe for amygdala function, was carried out in 97 children aged 9–14 years including 48 (12 female) with ASD and 49 (14 female) with typical development (TD). In addition, exploratory analyses were conducted examining the association between FPS and amygdala volume as assessed with magnetic resonance imaging in a subset of the children with ASD with or without an anxiety disorder with available MRI data. While the startle latency was increased in the children with ASD, there was no group difference in FPS. FPS was not significantly associated with traditional Diagnostic and Statistical Manual (DSM) or “autism distinct” forms of anxiety. Within the autism group, FPS was negatively correlated with amygdala volume. Multiple regression analyses revealed that the association between FPS and anxiety severity was significantly moderated by the size of the amygdala, such that the association between FPS and anxiety was significantly more positive in children with larger amygdalas than smaller amygdalas. These findings highlight the heterogeneity of emotional reactivity associated with ASD and the difficulties in establishing biologically meaningful probes of altered brain function. Lay summary Many children with autism spectrum disorder (ASD) have additional problems such as anxiety that can greatly impact their lives. How these co‐occurring symptoms develop is not well understood. We studied the amygdala, a region of the brain critical for processing fear and a laboratory method called fear potentiated startle for measuring fear conditioning, in children with ASD (with and without an anxiety disorder) and typically developing children. Results showed that the connection between fear conditioning and anxiety is dependent on the size of the amygdala in children with ASD.
... DTIStudio (www.mristudio.org) was used to align all DWIs to the average of b 0 s to remove small sample displacements due to vibrations during the long scan and compute average DWI (aDWI) and fractional anisotropy (FA) maps using the diffusion tensor model. The aDWI and FA maps were normalized to an MRI-based atlas (Chuang et al., 2011, Arefin et al., 2019 using the dual-channel (aDWI+FA) large deformation diffeomorphic metric mapping (LDDMM) (Ceritoglu et al., 2009). To assess global and regional amygdala connectivity, we transferred 14 structural labels (i.e. 7 structures in each hemisphere, e.g., cortex, hippocampus, cerebellum, etc. as defined in the Allen mouse brain atlas) to the subject images using the inverse mapping from LDDMM. ...
... The atlas was registered from the template to each subject space using a non-linear large deformation diffeomorphic metric mapping transformation with dual contrast (fractional anisotropy (FA) and b0). 8 This atlas-based approach has several advantages over voxel-based or white matter skeleton-based (e.g. tract-based spatial statistics 9 ) analytical methods by capturing the mean value from all the voxels in an ROI, improving measurement accuracy and increasing the effect size to detect changes. ...
Article
Background and purposes: Minimal hepatic encephalopathy (MHE) has no recognizable clinical symptoms, but patients have cognitive and psychomotor deficits. Hyperammonemia along with neuroinflammation lead to microstructural changes in cerebral parenchyma. Changes at conventional imaging are detected usually at the overt clinical stage, but microstruc- tural alterations by advanced magnetic resonance imaging techniques can be detected at an early stage. Materials and methods: Whole brain diffusion kurtosis imaging (DKI) data acquired at 3T was analyzed to investigate microstructural parenchymal changes in 15 patients with MHE and compared with 15 age- and sex-matched controls. DKI parametric maps, namely kurtosis fractional anisotropy (kFA), mean kurtosis (MK), axial kurtosis (AK) and radial kurtosis (RK), were evaluated at 64 white matter (WM) and gray matter (GM) regions of interest (ROIs) in the whole brain and correlated with the psychometric hepatic encephalopathy score (PHES). Results: The MHE group showed a decrease in kFA and AK across the whole brain, whereas MK and RK decreased in WM ROIs but increased in several cortical and deep GM ROIs. These alterations were consistent with brain regions involved in cognitive function. Significant moderate to strong correlations (–0.52 to –0.66; 0.56) between RK, MK and kFA kurtosis metrics and PHES were observed. Conclusion: DKI parameters show extensive microstructural brain abnormalities in MHE with minor correlation between the severity of tissue damage and psychometric scores.
... Gray and white matter labels de ned by the LONI Probabilistic Brain Atlas protocol (LPBA40) (33,34) were segmented using multi-atlas image segmentation. Multi-atlas image segmentation registers a set of age-appropriate anatomically labeled atlases onto a target brain image using diffeomorphic registration, which produces a set of candidate segmentations, and consensus segmentations are then produced using joint-label fusion, an advanced voting procedure (35)(36)(37). ...
Preprint
Full-text available
Background: Intellectual disability affects approximately one third of individuals with autism spectrum disorder (autism), yet a major unresolved question remains concerning the neurobiology that differentiates autistic individuals with and without intellectual disability. IQ is highly variable during childhood. We previously identified subgroups of autistic children with different trajectories of intellectual development from early childhood (2-3½ yeas) up to middle childhood (9-12 years): (a) Persistently-High: Individuals whose intelligence quotients (IQs) remained in the average or better range during this period, (b) Persistently-Low: Individuals whose IQs remained in the range of intellectual disability (IQ < 70) throughout development, and (c) Changers: Individuals whose IQs began in the range of intellectual disability but increased to the borderline or normal IQ range by middle childhood. In the present research, we sought to identify neurobiology that differentiates these trajectory-defined groups within our autism cohort in two brain networks with established links to intellectual functioning and its impairment in (1) the frontoparietal network (FPN), and (2) the default mode network (DMN). Methods: We conducted multivariate distance matrix regression (MDMR) and effect size analyses to examine the volumes of 22 brain regions (11 regions x 2 hemispheres) within the FPN and 24 (12 regions x 2 hemispheres) within the DMN in 48 Persistently-High (18 female), 108 Persistently-Low (32 female), and 109 Changers (39 female) using structural MRI that had been acquired at baseline, and IQ measurements from up to three time points spanning early to middle childhood (Mean Age Time 1: 3.2 years; Time 2: 5.4 years; Time 3: 11.3 years). FPN and DMN network regions of interest were defined on the basis of the large-scale networks defined in Smith et al., (2009). Results: Changers exhibited different DMN network structure from both Persistently-Low and Persistently-High trajectory groups at baseline, but the Persistently-High did not differ from the Persistently-Low group, suggesting that DMN structure may be an early predictor for change in IQ trajectory across childhood. In contrast, Persistently-High exhibited differences in the FPN from both Persistently-Low and Changers groups at baseline, suggesting a difference related more to concurrent IQ and the absence of intellectual disability. Conclusions: Within autism, DMN structure at baseline may differentiate individuals with persistently low IQ from those with more transitory low IQ that improves to the borderline range or better through early childhood, potentially indicating compensatory mechanisms which may be targeted by future interventions. The brain structure differences between these three IQ-based subgroups may be indicative of distinct neural underpinnings of autism phenotypic subtypes.
Article
Full-text available
Translational work in rodents elucidates basic mechanisms that drive complex behaviors relevant to psychiatric and neurological conditions. Nonetheless, numerous promising studies in rodents later fail in clinical trials, highlighting the need for improving the translational utility of preclinical studies in rodents. Imaging of small rodents provides an important strategy to address this challenge, as it enables a whole-brain unbiased search for structural and dynamic changes that can be directly compared to human imaging. The functional significance of structural changes identified using imaging can then be further investigated using molecular and genetic tools available for the mouse. Here, we describe a pipeline for unbiased search and characterization of structural changes and network properties, based on diffusion MRI data covering the entire mouse brain at an isotropic resolution of 100 µm. We first used unbiased whole-brain voxel-based analyses to identify volumetric and microstructural alterations in the brain of adult mice exposed to unpredictable postnatal stress (UPS), which is a mouse model of complex early life stress (ELS). Brain regions showing structural abnormalities were used as nodes to generate a grid for assessing structural connectivity and network properties based on graph theory. The technique described here can be broadly applied to understand brain connectivity in other mouse models of human disorders, as well as in genetically modified mouse strains. Graphic abstract: Pipeline for characterizing structural connectome in the mouse brain using diffusion magnetic resonance imaging. Scale bar = 1 mm.
Article
Vertebrate brains commonly consist of five basic embryologic anatomical regions: telencephalon; diencephalon; mesencephalon; metencephalon; and myelencephalon. The proportions of these regions vary widely across species and developmental stages. Investigation of their growth trajectories, therefore, has the potential to provide an understanding of the substrates of inter-species variation in neuroanatomy and function. To investigate the volumetric growth trajectories, structural magnetic resonance imaging (MRI) scans obtained from 618 healthy children (334 boys, 284 girls; ages 3-17 years old) were parcellated into five regions for the volume quantification. The sex- and region-specific growth trajectories were identified, and the most active growth was seen in the mesencephalon for both boys and girls. Whether similar regional growth patterns are seen in other species, or whether such patterns are related to evolution, are important questions that must be elucidated in the future.
Article
Knowing the volumetric changes in brain can allow for the estimation of the disease progression of various neurodegenerative disorders. Many studies have been shown that the volumetric changes in the some brain structures especially including the dopaminergic neurons, in patients with Parkinson’s disease (PD). The objective of this study was to compare intracerebral ventricles volume in patients with PD and healthy subjects to compare an automated atlas-based method (MRIStudio software) and a manual method (ImageJ). T1-weighted brain Magnetic Resonance Imaging (MRI) data of 21 patients with PD and 20 healthy individuals were used to calculate the intracerebral ventricle volumes. Measurement results obtained by ImageJ were considered as the gold standard. We found a significant increase in the left occipital part of the lateral ventricle volume in the patients with PD compared to the control subjects (p < 0.05). Also, no significant difference was found between the two methods of measurement (p > 0.05), meaning that a substantial agreement was found between the results obtained with the atlas-based analysis and manual method. The present study showed that MRIStudio can be performed easily and accurately on routine MRI scans for which the total intracerebral ventricles volume is to be estimated in PD. We suggest that, the attained volume values of intracerebral ventricles may provide a precious data for volumetric dependences of the anatomical structures in several clinical conditions.
Article
Classical deformable registration techniques achieve impressive results and offer a rigorous theoretical treatment, but are computationally intensive since they solve an optimization problem for each image pair. Recently, learning-based methods have facilitated fast registration by learning spatial deformation functions. However, these approaches use restricted deformation models, require supervised labels, or do not guarantee a diffeomorphic (topology-preserving) registration. Furthermore, learning-based registration tools have not been derived from a probabilistic framework that can offer uncertainty estimates. In this paper, we build a connection between classical and learning-based methods. We present a probabilistic generative model and derive an unsupervised learning-based inference algorithm that uses insights from classical registration methods and makes use of recent developments in convolutional neural networks (CNNs). We demonstrate our method on a 3D brain registration task for both images and anatomical surfaces, and provide extensive empirical analyses of the algorithm. Our principled approach results in state of the art accuracy and very fast runtimes, while providing diffeomorphic guarantees. Our implementation is available online at http://voxelmorph.csail.mit.edu.
Article
Quantification of tissue magnetic susceptibility using MRI offers a non-invasive measure of important tissue components in the brain, such as iron and myelin, potentially providing valuable information about normal and pathological conditions during aging. Despite many advances made in recent years on imaging techniques of quantitative susceptibility mapping (QSM), accurate and robust automated segmentation tools for QSM images that can help generate universal and sharable susceptibility measures in a biologically meaningful set of structures are still not widely available. In the present study, we developed an automated process to segment brain nuclei and quantify tissue susceptibility in these regions based on a susceptibility multi-atlas library, consisting of 10 atlases with T1-weighted images, gradient echo (GRE) magnitude images and QSM images of brains with different anatomic patterns. For each atlas in this library, 10 regions of interest in iron-rich deep gray matter structures that are better defined by QSM contrast were manually labeled, including caudate, putamen, globus pallidus internal/external, thalamus, pulvinar, subthalamic nucleus, substantia nigra, red nucleus and dentate nucleus in both left and right hemispheres. We then tested different pipelines using different combinations of contrast channels to bring the set of labels from the multi-atlases to each target brain and compared them with the gold standard manual delineation. The results showed that the segmentation accuracy using dual contrasts QSM/T1 pipeline outperformed other dual-contrast or single-contrast pipelines. The dice values of 0.77 ± 0.09 using the QSM/T1 multi-atlas pipeline rivaled with the segmentation reliability obtained from multiple evaluators with dice values of 0.79 ± 0.07 and gave comparable or superior performance in segmenting subcortical nuclei in comparison with standard FSL FIRST or recent multi-atlas package of volBrain. The segmentation performance of the QSM/T1 multi-atlas was further tested on QSM images acquired using different acquisition protocols and platforms and showed good reliability and reproducibility with average dice of 0.79 ± 0.08 to manual labels and 0.89 ± 0.04 in an inter-protocol manner. The extracted quantitative magnetic susceptibility values in the deep gray matter nuclei also correlated well between different protocols with inter-protocol correlation constants all larger than 0.97. Such reliability and performance was ultimately validated in an external dataset acquired at another study site with consistent susceptibility measures obtained using the QSM/T1 multi-atlas approach in comparison to those using manual delineation. In summary, we designed a susceptibility multi-atlas tool for automated and reliable segmentation of QSM images and for quantification of magnetic susceptibilities. It is publicly available through our cloud-based platform (www.mricloud.org). Further improvement on the performance of this multi-atlas tool is expected by increasing the number of atlases in the future.
Preprint
Full-text available
Purpose: The XCAT phantom allows for highly sophisticated multimodality imaging research. It includes a complete set of organs, muscle, bone, soft tissue, while also accounting for age, sex, and body mass index (BMI), which allows phantom studies to be performed at a population scale. At the same time, the XCAT phantom does not currently include the lymphatic system, critical for evaluating bulky nodal malignancies in lymphoma. We aimed to incorporate a full lymphatic system into the XCAT phantom and to generate realistic simulated images via guidance from lymphoma patient studies. Methods: A template lymphatic system model was used to define 276 lymph nodes and corresponding vessels using non-uniform rational basis spline (NURBS) surfaces. Lymph node properties were modified using the Rhinoceros 3D viewing software. The XCAT general parameter script was used to input organ concentrations and generate binary files with uptake and attenuation information. Results: Lymph nodes can be scaled, asymmetrically stretched, and translated within the intuitive Rhinoceros interface, to allow for realistic simulation of different lymph node pathologies. Bulky lymphoma tumours were generated in the mediastinum using expanded lymph nodes. Our results suggest that optimized thresholding (20-25%) provides better accuracy for determining total metabolic tumour volume (TMTV) of lymphoma tumours, while the gradient method was most accurate for total lesion glycolysis (TLG). Conclusions: An upgraded XCAT phantom with a fully simulated lymphatic system was created. Distributed to the research community, the XCAT phantom with the new lymphatic system has the potential of enabling studies to optimize image quality and quantitation, towards improved assessment of lymphoma including predictive modeling (e.g. improved TMTV and radiomics research).
Article
There has been a spurt in structural neuroimaging studies of the effect of hearing loss on the brain. Specifically, magnetic resonance imaging (MRI) and diffusion tensor imaging (DTI) technologies provide an opportunity to quantify changes in gray and white matter structures at the macroscopic scale. To date, there have been 32 MRI and 23 DTI studies that have analyzed structural differences accruing from pre- or peri-lingual pediatric hearing loss with congenital or early onset etiology and postlingual hearing loss in pre-to-late adolescence. Additionally, there have been 15 prospective clinical structural neuroimaging studies of children and adolescents being evaluated for cochlear implants. The results of the 70 studies are summarized in two figures and three tables. Plastic changes in the brain are seen to be multifocal rather than diffuse, that is, differences are consistent across regions implicated in the hearing, speech and language networks regardless of modes of communication and amplification. Structures in that play an important role in cognition are affected to a lesser extent. A limitation of these studies is the emphasis on volumetric measures and on homogeneous groups of subjects with hearing loss. It is suggested that additional measures of morphometry and connectivity could contribute to a greater understanding of the effect of hearing loss on the brain. Then an interpretation of the observed macroscopic structural differences is given. This is followed by discussion of how structural imaging can be combined with functional imaging to provide biomarkers for longitudinal tracking of amplification. This article is categorized under: Developmental Biology > Developmental Processes in Health and Disease Translational, Genomic, and Systems Medicine > Translational Medicine Laboratory Methods and Technologies > Imaging.
Article
Modest expansion of the human brain cerebrospinal fluid (CSF)-filled ventricles is normal with aging, and because of this, it can be difficult for physicians to accurately diagnose and treat enlarged ventricles (ventriculomegaly), called hydrocephalus1 (fluid or water in the brain) Ventriculomegaly occurs due to an obstruction (such as a blood clot or tumor), or a change in CSF absorption2. Primary hydrocephalus, also called idiopathic normal pressure hydrocephalus (iNPH), is non-obstructive and may be comorbid with other neurodegenerative diseases such as Alzheimer’s disease (AD) or frontotemporal dementia (FTD). Clinically, it can be difficult to tell whether the pathophysiological changes leading to cognitive impairment also led to the ventriculomegaly, as may occur in AD, versus whether the hydrocephalus itself is driving cognitive and motor impairment, i.e. iNPH. The goal of this thesis project is to investigate the relationship between iNPH and AD in order to better understand how they may contribute to each other, and to help clinicians distinguish between them. To do this, we compared cognitive performance and white matter integrity between patients with “pure” iNPH, “pure” Alzheimer’s disease (AD), and co-morbid iNPH + AD. Our results demonstrated that there are specific periventricular structures in the brain that are associated with cognitive impairment in AD versus iNPH. We conclude that the distribution pattern of AD vs. iNPH may be a valid tool to distinguish between these disorders, and may form the basis for subsequent studies that can further explicate the link between these often-overlapping etiologies.
Preprint
Full-text available
While the contributions of some genes to neuropsychiatric disorders are clear, the downstream neuronal effects are poorly understood. Over-expression of SHANK3 , which encodes a key synaptic protein, causes neuropsychiatric phenotypes in humans and manic-like behavior in mice providing an opportunity to interrogate the role of SHANK3 in a subset of neurons that might underlie the manic-like behavior. Herein, we describe Shank3’s critical role in D2 dopamine receptor (D2dr) neurons and show that Shank3 overexpression causes increased synaptic neurotransmission in D2dr, but not D1dr, expressing striatal medium spiny neurons. Either pharmacologic D2dr inhibition or genetic normalization of Shank3 abundance in D2-neurons ameliorates manic-like behaviors. Integrating bioinformatic analyses of Shank3’s striatal interactome, D1 and D2 dopamine receptor binding proteins, and single-cell RNA-seq datasets, we demonstrate a functional relationship between Shank3 and the D2dr—but not the D1dr. Thus, while Shank3 is over-expressed in both D1 and D2 dopamine receptor expressing striatal neurons, D2 neuronal dysfunction causes manic-like behaviors.
Article
Full-text available
Neurite orientation dispersion and density imaging (NODDI) is a diffusion model specifically designed for brain magnetic resonance imaging. Despite recent studies suggesting that NODDI modeling might be more sensitive to brain development than diffusion tensor imaging (DTI), these studies were limited to a relatively small age range and mainly based on the manually operated region of interest analysis. Therefore, this study applied NODDI to investigate brain development in a large sample size of 214 subjects ranging in ages from 0 to 14. The whole brain was automatically segmented into 122 regions. The maturation trajectory of each region was characterized by the time course of diffusion metrics and further quantified using nonlinear regression. The NODDI-derived metrics, neurite density index (NDI) and orientation dispersion index (ODI), increased with age. And these two metrics were superior to the DTI-derived metrics in SVM regression models of age. The NDI in white matter exhibited a more rapid growth than that in gray matter (including the cortex and deep nucleus). These diffusion indicators experienced conspicuous increases during early childhood and the growth speed slowed down in adolescence. Region-specific maturation patterns were described throughout the brain, including white matter, cortical and deep gray matter. These development patterns were evaluated and discussed on the basis of NODDI’s model assumptions. To summarize, this study verified the high sensitivity of NODDI to age over a crucial developmental period from newborn to adolescence. Moreover, the existing knowledge of brain development has been complemented, suggesting that NODDI has a potential capability in the investigation of brain development.
Article
Image registration is a fundamental task in medical image analysis. Recently, many deep learning based image registration methods have been extensively investigated due to their comparable performance with the state-of-the-art classical approaches despite the ultra-fast computational time. However, the existing deep learning methods still have limitations in the preservation of original topology during the deformation with registration vector fields. To address this issues, here we present a cycle-consistent deformable image registration, dubbed CycleMorph. The cycle consistency enhances image registration performance by providing an implicit regularization to preserve topology during the deformation. The proposed method is so flexible that can be applied for both 2D and 3D registration problems for various applications, and can be easily extended to multi-scale implementation to deal with the memory issues in large volume registration. Experimental results on various datasets from medical and non-medical applications demonstrate that the proposed method provides effective and accurate registration on diverse image pairs within a few seconds. Qualitative and quantitative evaluations on deformation fields also verify the effectiveness of the cycle consistency of the proposed method.
Article
Full-text available
Diffusion MRI (dMRI) is commonly used to map large axonal pathways in the white matter. Recent technical advances have also enabled dMRI to resolve the small and complex axonal and dendritic projections in the gray matter. This study investigated whether high-resolution dMRI can resolve the hippocampal neuronal projections and detect abnormal connections due to neurological injury. We performed 3D high spatial and angular resolution dMRI of the mouse brains of the offspring survivors from a model of intrauterine (UI) inflammation, who had known functional deficiency in the hippocampus. We used a novel hippocampal connection mapping method to quantify the intra- and inter-hippocampal projections among 34 automatically segmented hippocampal sub-regions. The results demonstrated wide-spread intra-hippocampal projections, but rather specific intra-hippocampal projections that primarily connected through the CA3 region. Compared with the control group (n = 9), UI-injured mice (n = 11) exhibited significantly reduced inter-hippocampal projection strength (p < 0.01), which correlated well with the neurobehavioral assessments (R2 = 0.47). Furthermore, using a whole-brain fixel-based analysis, we identified reduced fiber-density in the CA3 and the ventral hippocampal commissure of the UI-injured mice, which may explain the reduced inter-hippocampal projections. Histological findings also indicated reduced commissural fibers due to the UI-injury. Our study suggested that the dMRI-based connectivity mapping technique can potentially characterize abnormal hippocampal projections in neurological disorders.
Article
OBJECTIVE Comparative evaluation of diffusion kurtosis imaging (DKI) and diffusion tensor imaging (DTI) using a whole-brain atlas to comprehensively evaluate microstructural changes in the brain of Alzheimer disease (AzD) patients. METHODS Twenty-seven AzD patients and 25 age-matched controls were included. MRI data was analyzed using a whole-brain atlas with inclusion of 98 region of interests. White matter (WM) microstructural changes were assessed by Fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), radial diffusivity (RD), Kurtosis fractional anisotropy (KFA), mean kurtosis (MK), axial kurtosis (AK) and radial kurtosis (RK). Gray matter (GM) integrity was evaluated using KFA, MK, RK, AK and MD. Comparison of the DKI and DTI metrics were done using student t-test (p ≤ 0.001). RESULTS In AzD patients widespread increase in MD, AD and RD were found in various WM and GM region of interests. The extent of abnormality for DKI parameters was more limited in both GM and WM regions and revealed reduced kurtosis values except in lentiform nuclei. Both DKI and DTI parameters were sensitive to detect abnormality in WM areas with coherent and complex fiber arrangement. Receiver operating characteristic curve analysis for hippocampal values revealed the highest specificity of 88% for AK <0.6965 and highest sensitivity of 95.2% for MD >1.2659. CONCLUSION AzD patients have microstructural changes in both WM and GM and are well-depicted by both DKI and DTI. The alterations in kurtosis parameters, however, are more limited and correlate with areas in the brain primarily involved in cognition.
Article
Full-text available
Radial glial progenitor cells (RGPs) are the major neural progenitor cells that generate neurons and glia in the developing mammalian cerebral cortex1,2,3,4. In RGPs, the centrosome is positioned away from the nucleus at the apical surface of the ventricular zone of the cerebral cortex5,6,7,8. However, the molecular basis and precise function of this distinctive subcellular organization of the centrosome are largely unknown. Here we show in mice that anchoring of the centrosome to the apical membrane controls the mechanical properties of cortical RGPs, and consequently their mitotic behaviour and the size and formation of the cortex. The mother centriole in RGPs develops distal appendages that anchor it to the apical membrane. Selective removal of centrosomal protein 83 (CEP83) eliminates these distal appendages and disrupts the anchorage of the centrosome to the apical membrane, resulting in the disorganization of microtubules and stretching and stiffening of the apical membrane. The elimination of CEP83 also activates the mechanically sensitive yes-associated protein (YAP) and promotes the excessive proliferation of RGPs, together with a subsequent overproduction of intermediate progenitor cells, which leads to the formation of an enlarged cortex with abnormal folding. Simultaneous elimination of YAP suppresses the cortical enlargement and folding that is induced by the removal of CEP83. Together, these results indicate a previously unknown role of the centrosome in regulating the mechanical features of neural progenitor cells and the size and configuration of the mammalian cerebral cortex.
Preprint
Full-text available
Loss of hand function after stroke is a major cause of long-term disability. Hand function can be partitioned into strength and independent control of fingers (individuation). Here we developed a novel paradigm, which independently quantifies these two aspects of hand function, to track hand recovery in 54 patients with hemiparesis over the first year after their stroke. Most recovery of both strength and individuation occurred in the first three months after stroke. Improvement in strength and individuation were tightly correlated up to a strength level of approximately 60% of the unaffected side. Beyond this threshold, further gains in strength were not accompanied by improvements in individuation. Any observed improvements in individuation beyond the 60% threshold were attributable instead to a second independent stable factor. Lesion analysis revealed that damage to the hand area in motor cortex and the corticospinal tract (CST) correlated more with individuation than with strength. CST involvement correlated with individuation even after factoring out the strength-individuation correlation. The most parsimonious explanation for these behavioral and lesion-based findings is that most strength recovery, along with some individuation, can be attributed to descending systems other than the CST, whereas further recovery of individuation is CST dependent.
Article
Whole brain neuroanatomy using tera‐voxel light‐microscopic data sets is of much current interest. A fundamental problem in this field is the mapping of individual brain datasets to a reference space. Previous work has not rigorously quantified in‐vivo to ex‐vivo distortions in brain geometry from tissue processing. Further, existing approaches focus on registering uni‐modal volumetric data; however, given the increasing interest in the marmoset model for neuroscience research and the importance of addressing individual brain architecture variations, new algorithms are necessary to cross‐register multimodal datasets including MRIs and multiple histological series. Here we present a computational approach for same‐subject multimodal MRI‐guided reconstruction of a series of consecutive histological sections, jointly with diffeomorphic mapping to a reference atlas. We quantify the scale change during different stages of brain histological processing using the Jacobian determinant of the diffeomorphic transformations involved. By mapping the final image stacks to the ex‐vivo post‐fixation MRI, we show that a) tape‐transfer assisted histological sections can be re‐assembled accurately into 3D volumes with a local scale change of 2.0 ± 0.4% per axis dimension; in contrast, b) tissue perfusion/fixation as assessed by mapping the in‐vivo MRIs to the ex‐vivo post fixation MRIs shows a larger median absolute scale change of 6.9 ± 2.1% per axis dimension. This is the first systematic quantification of local metric distortions associated with whole‐brain histological processing, and we expect that the results will generalize to other species. These local scale changes will be important for computing local properties to create reference brain maps. This article is protected by copyright. All rights reserved.
Article
We present VoxelMorph, a fast learning-based framework for deformable, pairwise medical image registration. Traditional registration methods optimize an objective function for each pair of images, which can be time-consuming for large datasets or rich deformation models. In contrast to this approach and building on recent learning-based methods, we formulate registration as a function that maps an input image pair to a deformation field that aligns these images. We parameterize the function via a convolutional neural network and optimize the parameters of the neural network on a set of images. Given a new pair of scans, VoxelMorph rapidly computes a deformation field by directly evaluating the function. In this paper, we explore two different training strategies. In the first (unsupervised) setting, we train the model to maximize standard image matching objective functions that are based on the image intensities. In the second setting, we leverage auxiliary segmentations available in the training data. We demonstrate that the unsupervised model’s accuracy is comparable to the state-of-the-art methods while operating orders of magnitude faster. We also show that VoxelMorph trained with auxiliary data improves registration accuracy at test time and evaluate the effect of training set size on registration. Our method promises to speed up medical image analysis and processing pipelines while facilitating novel directions in learning-based registration and its applications. Our code is freely available at https://github.com/voxelmorph/voxelmorph .
Article
Full-text available
This paper examine the Euler-Lagrange equations for the solution of the large deformation diffeomorphic metric mapping problem studied in Dupuis et al. (1998) and Trouvé (1995) in which two images I 0, I 1 are given and connected via the diffeomorphic change of coordinates I 0○ϕ−1=I 1 where ϕ=Φ1 is the end point at t= 1 of curve Φ t , t∈[0, 1] satisfying .Φ t =v t (Φ t ), t∈ [0,1] with Φ0=id. The variational problem takes the form \mathop {\arg {\text{m}}in}\limits_{\upsilon :\dot \phi _t = \upsilon _t \left( {\dot \phi } \right)} \left( {\int_0^1 {\left\| {\upsilon _t } \right\|} ^2 {\text{d}}t + \left\| {I_0 \circ \phi _1^{ - 1} - I_1 } \right\|_{L^2 }^2 } \right)
Article
Full-text available
Mathematical techniques are presented for the transformation of digital anatomical textbooks from the ideal to the individual, allowing for the representation of the variabilities manifest in normal human anatomies. The ideal textbook is constructed on a fixed coordinate system to contain all of the information currently available about the physical properties of neuroanatomies. This information is obtained via sensor probes such as magnetic resonance, as well as computed axial and emission tomography, along with symbolic information such as white- and gray-matter tracts, nuclei, etc. Human variability associated with individuals is accommodated by defining probabilistic transformations on the textbook coordinate system, the transformations forming mathematical translation groups of high dimension. The ideal is applied to the individual patient by finding the transformation which is consistent with physical properties of deformable elastic solids and which brings the coordinate system of the textbook to that of the patient. Registration, segmentation, and fusion all result automatically because the textbook carries symbolic values as well as multisensor features.
Article
Full-text available
This paper proposes a method to match diffusion tensor magnetic resonance images (DT-MRIs) through the large deformation diffeomorphic metric mapping of vector fields, focusing on the fiber orientations, considered as unit vector fields on the image volume. We study a suitable action of diffeomorphisms on such vector fields, and provide an extension of the Large Deformation Diffeomorphic Metric Mapping framework to this type of dataset, resulting in optimizing for geodesics on the space of diffeomorphisms connecting two images. Existence of the minimizers under smoothness assumptions on the compared vector fields is proved, and coarse to fine hierarchical strategies are detailed, to reduce both ambiguities and computation load. This is illustrated by numerical experiments on DT-MRI heart images.
Article
New theoretical and practical concepts are presented for considerably enhancing the performance of magnetic resonance imaging (MRI) by means of arrays of multiple receiver coils. Sensitivity encoding (SENSE) is based on the fact that receiver sensitivity generally has an encoding effect complementary to Fourier preparation by linear field gradients. Thus, by using multiple receiver coils in parallel scan time in Fourier imaging can be considerably reduced. The problem of image reconstruction from sensitivity encoded data is formulated in a general fashion and solved for arbitrary coil configurations and k-space sampling patterns. Special attention is given to the currently most practical case, namely, sampling a common Cartesian grid with reduced density. For this case the feasibility of the proposed methods was verified both in vitro and in vivo. Scan time was reduced to one-half using a two-coil array in brain imaging. With an array of five coils double-oblique heart images were obtained in one-third of conventional scan time. Magn Reson Med 42:952-962, 1999.
Article
Diffusion tensor magnetic resonance imaging (DT-MRI) is unique in providing information about both the structural integrity and the orientation of white matter fibers in vivo and, through “tractography”, revealing the trajectories of white matter tracts. DT-MRI is therefore a promising technique for detecting differences in white matter architecture between different subject populations. However, while studies involving analyses of group averages of scalar quantities derived from DT-MRI data have been performed, as yet there have been no similar studies involving the whole tensor. Here we present the first step towards realizing such a study, i.e., the spatial normalization of whole tensor data sets. The approach is illustrated by spatial normalization of 10 DT-MRI data sets to a standard anatomical template. Both qualitative and quantitative approaches are described for assessing the results of spatial normalization. Techniques are then described for combining the spatially normalized data sets according to three definitions of average, i.e., the mean, median, and mode of a distribution of tensors. The current absence of, and hence need for, appropriate statistical tests for comparison of results derived from group-averaged DT-MRI data sets is then discussed. Finally, the feasibility of performing tractography on the group-averaged DT-MRI data set is investigated and the possibility and implications of generating a generic map of brain connectivity from a group of subjects is considered.
Article
This paper studies mathematical methods in the emerging new discipline of Computational Anatomy. Herein we formalize the Brown/Washington University model of anatomyfollowing the global pattern theory introduced in [1, 2], in which anatomies are represented as deformable templates, collections of 0 � 1 � 2 � 3;dimensional manifolds. Typical structure is carried by the template with the variabilities accommodated via the application of random transformations to the background manifolds. The anatomical model is a quadruple ( � H � I � P), the background space = [ M of 0 � 1 � 2 � 3;dimensional manifolds, the set of di eomorphic transformations on the background space H: \$ , the space of idealized medical imagery I, and P the family of probability measures on H. The group of di eomorphic transformations H is chosen to be rich enough so that a large family of shapes may be generated with the topologies of the template maintained. For normal anatomy one deformable template is studied, with ( � H � I) corresponding to a homogeneous space [3], in that it can be completely generated from one of its elements, I = HItemp�Itemp 2I. For disease, a family of templates [ Itemp are introduced of perhaps varying dimensional transformation classes. The complete anatomy is is a collection of homogeneous spaces Itotal = [ (I � H). There are three principal components to computational anatomy studied herein.
Article
Several groups have developed methods for registering an individual's 3D MRI by deforming a standard template. This achievement leads to many possibilities for segmentation and morphology that will impact nuclear medical research in areas such as activation and receptor studies. Accordingly, there is a need for methods that can assess the accuracy of intersubject registration. We have developed a method based on a set of 128 anatomic landmarks per hemisphere, both cortical and subcortical, that allows assessment of both global and local transformation accuracy. We applied our method to compare the accuracy of two standard methods of intersubject registration, AIR 3.0 with fifth-order polynomial warping and the Talairach stereotaxic transformation (Talairach and Tournoux, 1988). SPGR MRI's (256 × 256 × 160) of six normal subjects (age 18–24 years) were derformed to match a standard template volume. To assess registration accuracy the landmarks were located on both the template volume and the transformed volumes by an experienced neuroanatomist. The resulting list of coordinates was analyzed graphically and by ANOVA to compare the accuracy of the two methods and the results of the manual analysis. ANOVA performed over all 128 landmarks showed that the Woods method was more accurate than Talairach (left hemisphereF = 2.8,P < 0.001 and right hemisphereF =2.4,P < 0.006). The Woods method provided a better brain surface transformation than did Talairach (F = 18.0,P < 0.0001), but as expected there was a smaller difference for subcortical structures and both had an accuracy <1 mm for the majority of subcortical landmarks. Overall, both the Woods and Talairach method located about 70% of landmarks with an error of 3 mm or less. More striking differences were noted for landmark accuracy ≤1 mm, where the Woods method located about 40% and Talairach about 23%. These results demonstrate that this anatomically based assessment method can help evaluate new methods of intersubject registration and should be a helpful tool in appreciating regional differences in accuracy. Consistent with expectation, we confirmed that the Woods nonlinear registration method was more accurate than Talairach. Landmark-based anatomic analyses of intersubject registration accuracy offer opportunities to explore the relationship among structure, function and architectonic boundaries in the human brain.
Article
At its simplest, voxel-based morphometry (VBM) involves a voxel-wise comparison of the local concentration of gray matter between two groups of subjects. The procedure is relatively straightforward and involves spatially normalizing high-resolution images from all the subjects in the study into the same stereotactic space. This is followed by segmenting the gray matter from the spatially normalized images and smoothing the gray-matter segments. Voxel-wise parametric statistical tests which compare the smoothed gray-matter images from the two groups are performed. Corrections for multiple comparisons are made using the theory of Gaussian random fields. This paper describes the steps involved in VBM, with particular emphasis on segmenting gray matter from MR images with nonuniformity artifact. We provide evaluations of the assumptions that underpin the method, including the accuracy of the segmentation and the assumptions made about the statistical distribution of the data.
Article
All fields of neuroscience that employ brain imaging need to communicate their results with reference to anatomical regions. In particular, comparative morphometry and group analysis of functional and physiological data require coregistration of brains to establish correspondences across brain structures. It is well established that linear registration of one brain to another is inadequate for aligning brain structures, so numerous algorithms have emerged to nonlinearly register brains to one another. This study is the largest evaluation of nonlinear deformation algorithms applied to brain image registration ever conducted. Fourteen algorithms from laboratories around the world are evaluated using 8 different error measures. More than 45,000 registrations between 80 manually labeled brains were performed by algorithms including: AIR, ANIMAL, ART, Diffeomorphic Demons, FNIRT, IRTK, JRD-fluid, ROMEO, SICLE, SyN, and four different SPM5 algorithms ("SPM2-type" and regular Normalization, Unified Segmentation, and the DARTEL Toolbox). All of these registrations were preceded by linear registration between the same image pairs using FLIRT. One of the most significant findings of this study is that the relative performances of the registration methods under comparison appear to be little affected by the choice of subject population, labeling protocol, and type of overlap measure. This is important because it suggests that the findings are generalizable to new subject populations that are labeled or evaluated using different labeling protocols. Furthermore, we ranked the 14 methods according to three completely independent analyses (permutation tests, one-way ANOVA tests, and indifference-zone ranking) and derived three almost identical top rankings of the methods. ART, SyN, IRTK, and SPM's DARTEL Toolbox gave the best results according to overlap and distance measures, with ART and SyN delivering the most consistently high accuracy across subjects and label sets. Updates will be published on the http://www.mindboggle.info/papers/ website.
Article
The diagonal and off-diagonal elements of the effective self-diffusion tensor, Deff, are related to the echo intensity in an NMR spin-echo experiment. This relationship is used to design experiments from which Deff is estimated. This estimate is validated using isotropic and anisotropic media, i.e., water and skeletal muscle. It is shown that significant errors are made in diffusion NMR spectroscopy and imaging of anisotropic skeletal muscle when off-diagonal elements of Deff are ignored, most notably the loss of information needed to determine fiber orientation. Estimation of Deff provides the theoretical basis for a new MRI modality, diffusion tensor imaging, which provides information about tissue microstructure and its physiologic state not contained in scalar quantities such as T1, T2, proton density, or the scalar apparent diffusion constant.
Article
To assess intrinsic properties of water diffusion in normal human brain by using quantitative parameters derived from the diffusion tensor, D, which are insensitive to patient orientation. Maps of the principal diffusivities of D, of Trace(D), and of diffusion anisotropy indices were calculated in eight healthy adults from 31 multisection, interleaved echo-planar diffusion-weighted images acquired in about 25 minutes. No statistically significant differences in Trace(D) (approximately 2,100 x 10(-6) mm2/sec) were found within normal brain parenchyma, except in the cortex, where Trace(D) was higher. Diffusion anisotropy varied widely among different white matter regions, reflecting differences in fiber-tract architecture. In the corpus callosum and pyramidal tracts, the ratio of parallel to perpendicular diffusivities was approximately threefold higher than previously reported, and diffusion appeared cylindrically symmetric. However, in other white matter regions, particularly in the centrum semiovale, diffusion anisotropy was low, and cylindrical symmetry was not observed. Maps of parameters derived from D were also used to segment tissues based on their diffusion properties. A quantitative characterization of water diffusion in anisotropic, heterogeneously oriented tissues is clinically feasible. This should improve the neuroradiologic assessment of a variety of gray and white matter disorders.
Article
Indices of diffusion anisotropy calculated from diffusion coefficients acquired in two or three perpendicular directions are rotationally variant. In living monkey brain, these indices severely underestimate the degree of diffusion anisotropy. New indices calculated from the entire diffusion tensor are rotationally invariant (RI). They show that anisotropy is highly variable in different white matter regions depending on the degree of coherence of fiber tract directions. In structures with a regular, parallel fiber arrangement, water diffusivity in the direction parallel to the fibers (Dparallel approximately 1400-1800 x 10(-6) mm2/s) is almost 10 times higher than the average diffusivity in directions perpendicular to them (D + D)/2 [corrected] approximately 150-300 x 10(-6) mm2/s), and is almost three times higher than previously reported. In structures where the fiber pattern is less coherent (e.g., where fiber bundles merge), diffusion anisotropy is significantly reduced. However, RI anisotropy indices are still susceptible to noise contamination. Monte Carlo simulations show that these indices are statistically biased, particularly those requiring sorting of the eigenvalues of the diffusion tensor based on their magnitude. A new intervoxel anisotropy index is proposed that locally averages inner products between diffusion tensors in neighboring voxels. This "lattice" RI index has an acceptably low error variance and is less susceptible to bias than any other RI anisotropy index proposed to date.
Article
This paper reviews recent developments by the Washington/Brown groups for the study of anatomical shape in the emerging new discipline of computational anatomy. Parametric representations of anatomical variation for computational anatomy are reviewed, restricted to the assumption of small deformations. The generation of covariance operators for probabilistic measures of anatomical variation on coordinatized submanifolds is formulated as an empirical procedure. Populations of brains are mapped to common coordinate systems, from which template coordinate systems are constructed which are closest to the population of anatomies in a minimum distance sense. Variation of several one-, two- and three-dimensional manifolds, i.e. sulci, surfaces and brain volumes are examined via Gaussian measures with mean and covariances estimated directly from maps of templates to targets. Methods are presented for estimating the covariances of vector fields from a family of empirically generated maps, posed as generalized spectrum estimation indexed over the submanifolds. Covariance estimation is made parametric, analogous to autoregressive modelling, by introducing small deformation linear operators for constraining the spectrum of the fields.
Article
The precise characterization of cortical connectivity is important for the understanding of brain morphological and functional organization. Such connectivity is conveyed by specific pathways or tracts in the white matter. Diffusion-weighted magnetic resonance imaging detects the diffusivity of water molecules in three dimensions. Diffusivity is anisotropic in oriented tissues such as fiber tracts. In the present study, we used this method to map (in terms of orientation, location, and size) the "stem" (compact portion) of the principal association, projection, and commissural white matter pathways of the human brain in vivo, in 3 normal subjects. In addition, its use in clinical neurology is illustrated in a patient with left inferior parietal lobule embolic infarction in whom a significant reduction in relative size of the stem of the left superior longitudinal fasciculus was observed. This represents an important method for the characterization of major association pathways in the living human that are not discernible by conventional magnetic resonance imaging. In the clinical domain, this method will have a potential impact on the understanding of the diseases that involve white matter such as stroke, multiple sclerosis, amyotrophic lateral sclerosis, head injury, and spinal cord injury.
Article
We sought to describe and validate an automated image registration method (AIR 3.0) based on matching of voxel intensities. Different cost functions, different minimization methods, and various sampling, smoothing, and editing strategies were compared. Internal consistency measures were used to place limits on registration accuracy for MRI data, and absolute accuracy was measured using a brain phantom for PET data. All strategies were consistent with subvoxel accuracy for intrasubject, intramodality registration. Estimated accuracy of registration of structural MRI images was in the 75 to 150 microns range. Sparse data sampling strategies reduced registration times to minutes with only modest loss of accuracy. The registration algorithm described is a robust and flexible tool that can be used to address a variety of image registration problems. Registration strategies can be tailored to meet different needs by optimizing tradeoffs between speed and accuracy.
Article
This paper investigates the use of color to represent the directional information contained in the diffusion tensor. Ideally, one wants to take into account both the properties of human color vision and of the given display hardware to produce a representation in which differences in the orientation of anisotropic structures are proportional to the perceived differences in color. It is argued here that such a goal cannot be achieved in general and therefore, empirical or heuristic schemes, which avoid some of the common artifacts of previously proposed approaches, are implemented. Directionally encoded color (DEC) maps of the human brain obtained using these schemes clearly show the main association, projection, and commissural white matter pathways. In the brainstem, motor and sensory pathways are easily identified and can be differentiated from the transverse pontine fibers and the cerebellar peduncles. DEC maps obtained from diffusion tensor imaging data provide a simple and effective way to visualize fiber direction, useful for investigating the structural anatomy of different organs. Magn Reson Med 42:526-540, 1999.
Article
New theoretical and practical concepts are presented for considerably enhancing the performance of magnetic resonance imaging (MRI) by means of arrays of multiple receiver coils. Sensitivity encoding (SENSE) is based on the fact that receiver sensitivity generally has an encoding effect complementary to Fourier preparation by linear field gradients. Thus, by using multiple receiver coils in parallel scan time in Fourier imaging can be considerably reduced. The problem of image reconstruction from sensitivity encoded data is formulated in a general fashion and solved for arbitrary coil configurations and k-space sampling patterns. Special attention is given to the currently most practical case, namely, sampling a common Cartesian grid with reduced density. For this case the feasibility of the proposed methods was verified both in vitro and in vivo. Scan time was reduced to one-half using a two-coil array in brain imaging. With an array of five coils double-oblique heart images were obtained in one-third of conventional scan time. Magn Reson Med 42:952-962, 1999.
Article
This paper investigates the use of color to represent the directional information contained in the diffusion tensor. Ideally, one wants to take into account both the properties of human color vision and of the given display hardware to produce a representation in which differences in the orientation of anisotropic structures are proportional to the perceived differences in color. It is argued here that such a goal cannot be achieved in general and therefore, empirical or heuristic schemes, which avoid some of the common artifacts of previously proposed approaches, are implemented. Directionally encoded color (DEC) maps of the human brain obtained using these schemes clearly show the main association, projection, and commissural white matter pathways. In the brainstem, motor and sensory pathways are easily identified and can be differentiated from the transverse pontine fibers and the cerebellar peduncles. DEC maps obtained from diffusion tensor imaging data provide a simple and effective way to visualize fiber direction, useful for investigating the structural anatomy of different organs. Magn Reson Med 42:526–540, 1999. © 1999 Wiley-Liss, Inc.
Article
John Ashburner and Karl Friston (2000) introduced a standardized method of "voxel-based morphometry" (VBM) for comparisons of local concentrations of gray matter between two groups of subjects. Segmented images of gray matter from grossly normalized high-resolution images are smoothed and their group differences analyzed by the now-conventional voxelwise Worsley approach to Gaussian random fields of differences. This comment concerns an unfortunate interaction between the algorithm's spatial normalization and voxelwise comparison steps, whereby several obvious quantitative confounds are injected at the core of the inference engine the authors put forward. Specifically, the statistics of the resulting voxelwise comparisons are uninformative about group differences wherever the spatial normalization algorithm has failed to register on any robustly appearing image gradient. The method of Ashburner and Friston is defensible only far from all image gradients.
Article
Diffusion tensor magnetic resonance imaging (DT-MRI) is unique in providing information about both the structural integrity and the orientation of white matter fibers in vivo and, through "tractography", revealing the trajectories of white matter tracts. DT-MRI is therefore a promising technique for detecting differences in white matter architecture between different subject populations. However, while studies involving analyses of group averages of scalar quantities derived from DT-MRI data have been performed, as yet there have been no similar studies involving the whole tensor. Here we present the first step towards realizing such a study, i.e., the spatial normalization of whole tensor data sets. The approach is illustrated by spatial normalization of 10 DT-MRI data sets to a standard anatomical template. Both qualitative and quantitative approaches are described for assessing the results of spatial normalization. Techniques are then described for combining the spatially normalized data sets according to three definitions of average, i.e., the mean, median, and mode of a distribution of tensors. The current absence of, and hence need for, appropriate statistical tests for comparison of results derived from group-averaged DT-MRI data sets is then discussed. Finally, the feasibility of performing tractography on the group-averaged DT-MRI data set is investigated and the possibility and implications of generating a generic map of brain connectivity from a group of subjects is considered.
Article
This work reports the use of diffusion tensor magnetic resonance tractography to visualize the three-dimensional (3D) structure of the major white matter fasciculi within living human brain. Specifically, we applied this technique to visualize in vivo (i) the superior longitudinal (arcuate) fasciculus, (ii) the inferior longitudinal fasciculus, (iii) the superior fronto-occipital (subcallosal) fasciculus, (iv) the inferior frontooccipital fasciculus, (v) the uncinate fasciculus, (vi) the cingulum, (vii) the anterior commissure, (viii) the corpus callosum, (ix) the internal capsule, and (x) the fornix. These fasciculi were first isolated and were then interactively displayed as a 3D-rendered object. The virtual tract maps obtained in vivo using this approach were faithful to the classical descriptions of white matter anatomy that have previously been documented in postmortem studies. Since we have been able to interactively delineate and visualize white matter fasciculi over their entire length in vivo, in a manner that has only previously been possible by histological means, "virtual in vivo interactive dissection" (VIVID) adds a new dimension to anatomical descriptions of the living human brain.
Article
A method for the spatial normalization and reorientation of diffusion tensor (DT) fields is presented. Spatial normalization of tensor fields requires an appropriate reorientation of the tensor on each voxel, in addition to its relocation into the standardized space. This appropriate tensor reorientation is determined from the spatial normalization transformation and from an estimate of the underlying fiber direction. The latter is obtained by treating the principal eigenvectors of the tensor field around each voxel as random samples drawn from the probability distribution that represents the direction of the underlying fiber. This approach was applied to DT images from nine normal volunteers, and the results show a significant improvement in signal-to-noise ratio (SNR) after spatial normalization and averaging of tensor fields across individuals. The statistics of the spatially normalized tensor field, which represents the tensor characteristics of normal individuals, may be useful for quantitatively characterizing individual variations of white matter structures revealed by DT imaging (DTI) and deviations caused by pathology. Simulated experiments using this methodology are also described.
Article
Two- and three-dimensional (3D) white matter atlases were created on the basis of high-spatial-resolution diffusion tensor magnetic resonance (MR) imaging and 3D tract reconstruction. The 3D trajectories of 17 prominent white matter tracts could be reconstructed and depicted. Tracts were superimposed on coregistered anatomic MR images to parcel the white matter. These parcellation maps were then compared with coregistered diffusion tensor imaging color maps to assign visible structures. The results showed (a). which anatomic structures can be identified on diffusion tensor images and (b). where these anatomic units are located at each section level and orientation. The atlas may prove useful for educational and clinical purposes.
Article
Diffusion Tensor MRI (DT-MRI) can provide important in vivo information for the detection of brain abnormalities in diseases characterized by compromised neural connectivity. To quantify diffusion tensor abnormalities based on voxel-based statistical analysis, spatial normalization is required to minimize the anatomical variability between studied brain structures. In this article, we used a multiple input channel registration algorithm based on a demons algorithm and evaluated the spatial normalization of diffusion tensor image in terms of the input information used for registration. Registration was performed on 16 DT-MRI data sets using different combinations of the channels, including a channel of T2-weighted intensity, a channel of the fractional anisotropy, a channel of the difference of the first and second eigenvalues, two channels of the fractional anisotropy and the trace of tensor, three channels of the eigenvalues of the tensor, and the six channel tensor components. To evaluate the registration of tensor data, we defined two similarity measures, i.e., the endpoint divergence and the mean square error, which we applied to the fiber bundles of target images and registered images at the same seed points in white matter segmentation. We also evaluated the tensor registration by examining the voxel-by-voxel alignment of tensors in a sample of 15 normalized DT-MRIs. In all evaluations, nonlinear warping using six independent tensor components as input channels showed the best performance in effectively normalizing the tract morphology and tensor orientation. We also present a nonlinear method for creating a group diffusion tensor atlas using the average tensor field and the average deformation field, which we believe is a better approach than a strict linear one for representing both tensor distribution and morphological distribution of the population.
Article
The objective of inter-subject registration of three-dimensional volumetric brain scans is to reduce the anatomical variability between the images scanned from different individuals. This is a necessary step in many different applications such as voxelwise group analysis of imaging data obtained from different individuals. In this paper, the ability of three different image registration algorithms in reducing inter-subject anatomical variability is quantitatively compared using a set of common high-resolution volumetric magnetic resonance imaging scans from 17 subjects. The algorithms are from the automatic image registration (AIR; version 5), the statistical parametric mapping (SPM99), and the automatic registration toolbox (ART) packages. The latter includes the implementation of a non-linear image registration algorithm, details of which are presented in this paper. The accuracy of registration is quantified in terms of two independent measures: (1) post-registration spatial dispersion of sets of homologous landmarks manually identified on images before or after registration; and (2) voxelwise image standard deviation maps computed within the set of images registered by each algorithm. Both measures showed that the ART algorithm is clearly superior to both AIR and SPM99 in reducing inter-subject anatomical variability. The spatial dispersion measure was found to be more sensitive when the landmarks were placed after image registration. The standard deviation measure was found sensitive to intensity normalization or the method of image interpolation.
Article
A versatile resource program was developed for diffusion tensor image (DTI) computation and fiber tracking. The software can read data formats from a variety of MR scanners. Tensor calculation is performed by solving an over-determined linear equation system using least square fitting. Various types of map data, such as tensor elements, eigenvalues, eigenvectors, diffusion anisotropy, diffusion constants, and color-coded orientations can be calculated. The results are visualized interactively in orthogonal views and in three-dimensional mode. Three-dimensional tract reconstruction is based on the Fiber Assignment by Continuous Tracking (FACT) algorithm and a brute-force reconstruction approach. To improve the time and memory efficiency, a rapid algorithm to perform the FACT is adopted. An index matrix for the fiber data is introduced to facilitate various types of fiber bundles selection based on approaches employing multiple regions of interest (ROIs). The program is developed using C++ and OpenGL on a Windows platform.
Article
In this paper, we present a novel deformable registration algorithm for diffusion tensor MR images that enables explicit optimization of tensor reorientation. The optimization seeks a piecewise affine transformation that divides the image domain into uniform regions and transform each region affinely. The objective function captures both the image similarity and the smoothness of the transformation across region boundaries. The image similarity enables explicit orientation optimization by incorporating tensor reorientation, which is necessary for warping diffusion tensor images. The objective function is formulated in a way that allows explicit implementation of analytic derivatives to drive fast and accurate optimization using the conjugate gradient method. By explicitly optimizing tensor reorientation, the algorithm is designed to take advantage of similarity measures comparing tensors as a whole. The optimal transformation is hierarchically refined in a subdivision framework. A comparison with affine registration for inter-subject normalization of 8 subjects shows that the proposed algorithm improves the alignment of several major white matter structures examined: the anterior thalamic radiations, the inferior fronto-occipital fasciculi, the corticospinal/corticobulbar tracts and the genu and the splenium of the corpus callosum. The alignment of white matter structures is assessed using a novel scheme of computing distances between the corresponding fiber bundles derived from tractography.
Article
White matter connectivity in the human brain can be mapped by diffusion tensor magnetic resonance imaging (DTI). After reconstruction, the diffusion tensors, the diffusion amplitude and the diffusion direction can be displayed on a morphological background. Consequently, diffusion tensor fibre tracking can be applied as a non-invasive in vivo technique for the delineation and quantification of specific white matter pathways. The aim of this study was to show that normalization to the Montreal Neurological Institute (MNI) stereotaxic standard space preserves specific diffusion features. Therefore, techniques for tensor imaging and fibre tracking were applied to the normalized brains as well as to the group averaged brain data. A normalization step of individual data was included by registration to a scanner- and sequence-specific DTI template data set which was created from a normal database transformed to MNI space. The algorithms were tested and validated for a group of 13 healthy controls.
Article
Brain registration to a stereotaxic atlas is an effective way to report anatomic locations of interest and to perform anatomic quantification. However, existing stereotaxic atlases lack comprehensive coordinate information about white matter structures. In this paper, white matter-specific atlases in stereotaxic coordinates are introduced. As a reference template, the widely used ICBM-152 was used. The atlas contains fiber orientation maps and hand-segmented white matter parcellation maps based on diffusion tensor imaging (DTI). Registration accuracy by linear and non-linear transformation was measured, and automated template-based white matter parcellation was tested. The results showed a high correlation between the manual ROI-based and the automated approaches for normal adult populations. The atlases are freely available and believed to be a useful resource as a target template and for automated parcellation methods.