[Show abstract][Hide abstract] ABSTRACT: In a continuous setting, diffeomorphisms generated by stationary velocity fields (SVF) are invertible transformations with differen-tiable inverses. However, due to the numerical integration of the velocity field, inverse consistency is not achieved in practice. In SVF based image registration, inverse consistency is therefore often enforced through a penalty term. Existing penalty terms penalize the inverse consistency error generated by the composition of the forward and backward transformations. However, in such terms, a higher consistency requirement pushes the transformation towards linearity due to the discretization involved and fixed number of integration time-steps. In this paper, we propose a method to both penalize inverse consistency error and to adaptively set the number of integration time-steps required, so that the predicted maximum inverse consistency error is bounded, taking into account discretization errors. This formulation allows more flexibility in the transformation model to realize complex deformations while still achieving the desired level of inverse consistency. Using synthetic examples, we show that the measured inverse consistency and the predicted inverse consistency match. Also, the proposed method is able to achieve more accurate image registration. On the MGH10 dataset, the Jaccard index of the proposed method on inter-subject registration reaches the same level as the registration scheme using a fixed-time step and the conventional penalty term while using a lower number of integration time-steps, thus saving on the computational time.
MICCAI Workshop on Mathematical Foundations of Computational Anatomy (MFCA); 10/2015
[Show abstract][Hide abstract] ABSTRACT: In this paper, we propose an automated Euler's time-step adjustment scheme for diffeomorphic image registration using stationary velocity fields (SVFs). The proposed variational problem aims at bounding the inverse consistency error by adaptively adjusting the number of Euler's step required to realize the time integration. This particular formulation allows us to gain computationally since only relevant number of time steps are taken. We parameterize the SVFs using multi-scale Wendland kernels through the kernel bundle framework. In terms of performance, the proposed scheme reaches the same accuracy as a fixed time-step scheme however at a much less computational cost.
International Symposium on Biomedical Imaging 2015; 07/2015
[Show abstract][Hide abstract] ABSTRACT: To investigate arterial spin labeling (ASL)-MRI for the early diagnosis of and differentiation between the two most common types of presenile dementia: Alzheimer's disease (AD) and frontotemporal dementia (FTD), and for distinguishing age-related from pathological perfusion changes.
Thirteen AD and 19 FTD patients, and 25 age-matched older and 22 younger controls underwent 3D pseudo-continuous ASL-MRI at 3 T. Gray matter (GM) volume and cerebral blood flow (CBF), corrected for partial volume effects, were quantified in the entire supratentorial cortex and in 10 GM regions. Sensitivity, specificity and diagnostic performance were evaluated in regions showing significant CBF differences between patient groups or between patients and older controls.
AD compared with FTD patients had hypoperfusion in the posterior cingulate cortex, differentiating these with a diagnostic performance of 74 %. Compared to older controls, FTD patients showed hypoperfusion in the anterior cingulate cortex, whereas AD patients showed a more widespread regional hypoperfusion as well as atrophy. Regional atrophy was not different between AD and FTD. Diagnostic performance of ASL to differentiate AD or FTD from controls was good (78-85 %). Older controls showed global hypoperfusion compared to young controls.
ASL-MRI contributes to early diagnosis of and differentiation between presenile AD and FTD.
• ASL-MRI facilitates differentiation of early Alzheimer's disease and frontotemporal dementia. • Posterior cingulate perfusion is lower in Alzheimer's disease than frontotemporal dementia. • Compared to controls, Alzheimer's disease patients show hypoperfusion in multiple regions. • Compared to controls, frontotemporal dementia patients show focal anterior cingulate hypoperfusion. • Global decreased perfusion in older adults differs from hypoperfusion in dementia.
European Radiology 05/2015; DOI:10.1007/s00330-015-3789-x · 4.01 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Computer-aided diagnosis of dementia using a support
vector machine (SVM) can be improved with feature
selection. The relevance of individual features can be quantified
from the SVM weights as a significance map (p-map). Although
these p-maps previously showed clusters of relevant voxels in
dementia-related brain regions, they have not yet been used
for feature selection. Therefore, we introduce two novel feature
selection methods based on p-maps using a direct approach
(filter) and an iterative approach (wrapper).
To evaluate these p-map feature selection methods, we compared
them with methods based on the SVM weight vector
directly, t-statistics and expert knowledge. We used MRI data
from the Alzheimer’s Disease Neuroimaging Initiative classifying
Alzheimer’s disease (AD) patients, mild cognitive impairment
(MCI) patients who converted to AD (MCIc), MCI patients who
did not convert to AD (MCInc), and cognitively normal controls
(CN). Features for each voxel were derived from gray matter
Feature selection based on the SVM weights gave better results
than t-statistics and expert knowledge. The p-map methods
performed slightly better than those using the weight vector.
The wrapper method scored better than the filter method.
Recursive feature elimination based on the p-map improved most
for AD-CN: the area under the receiver-operating-characteristic
curve (AUC) significantly increased from 90.3% without feature
selection to 92.0% when selecting 1.5%-3% of the features. This
feature selection method also improved the other classifications:
AD-MCI 0.1% improvement in AUC (not significant), MCI-CN
0.7%, and MCIc-MCInc 0.1% (not significant).
Although the performance improvement due to feature selection
was limited, the methods based on the p-map generally had
the best performance and were therefore better in estimating the
relevance of individual features.
IEEE Journal of Biomedical and Health Informatics 05/2015; 19(5). DOI:10.1109/JBHI.2015.2432832 · 1.44 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Algorithms for computer-aided diagnosis of dementia based on structural MRI have demonstrated high performance in the literature, but are difficult to compare as different data sets and methodology were used for evaluation. In addition, it is unclear how the algorithms would perform on previously unseen data, and thus, how they would perform in clinical practice when there is no real opportunity to adapt the algorithm to the data at hand. To address these comparability, generalizability and clinical applicability issues, we organized a grand challenge that aimed to objectively compare algorithms based on a clinically representative multi- center data set. Using clinical practice as starting point, the goal was to reproduce the clinical diagnosis. Therefore, we evaluated algorithms for multi-class classification of three diagnostic groups: patients with probable Alzheimer’s disease, patients with mild cognitive impairment and healthy controls. The diagnosis based on clinical criteria was used as reference standard, as it was the best available reference despite its known limitations. For evaluation, a previously unseen test set was used consisting of 354 T1-weighted MRI scans with the diagnoses blinded. Fifteen research teams participated with in total 29 algorithms. The algorithms were trained on a small training set (n=30) and optionally on data from other sources (e.g., the Alzheimer’s Disease Neuroimaging Initiative, the Australian Imaging Biomarkers and Lifestyle flagship study of aging). The best performing algorithm yielded an accuracy of 63.0% and an area under the receiver-operating-characteristic curve (AUC) of 78.8%. In general, the best performances were achieved using feature extraction based on voxel-based morphometry or a combination of features that included volume, cortical thickness, shape and intensity. The challenge is open for new submissions via the web-based framework: http://caddementia.grand-challenge.org.
[Show abstract][Hide abstract] ABSTRACT: In this work we present a framework for reliably detecting significant differences in quantitative magnetic resonance imaging (MRI) and evaluate it with diffusion tensor imaging (DTI) experiments. As part of this framework we propose a new spatially regularized maximum likelihood estimator that simultaneously estimates the quantitative parameters and the spatially -smoothly- varying noise level from the acquisitions. The noise level estimation method does not require repeated acquisitions. We show that the amount of regularization in this method can be set a-priori to achieve a desired coefficient of variation of the estimated noise level. The noise level estimate allows the construction of a Cram´er-Rao-lower-bound based test statistic that reliably assesses the significance of differences between voxels within a scan or across different scans. We show that the regularized noise level estimate improves upon existing methods and results in a substantially increased precision of the uncertainty estimates of the DTI parameters. It enables correct specification of the null distribution of the test statistic and with it the test statistic obtains the highest sensitivity and specificity. The source code of the estimation framework, test statistic and experiment scripts are made available to the community.
IEEE Transactions on Medical Imaging 12/2014; 34(5). DOI:10.1109/TMI.2014.2380830 · 3.39 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: In standard B-mode ultrasound (BMUS), segmentation of the lumen of atherosclerotic carotid arteries and studying the lumen geometry over time are difficult owing to irregular lumen shapes, noise, artifacts, and echolucent plaques. Contrast enhanced ultrasound (CEUS) improves lumen visualization, but lumen segmentation remains challenging owing to varying intensities, CEUS-specific artifacts and lack of tissue visualization. To overcome these challenges, we propose a novel method using simultaneously acquired BMUS&CEUS image sequences. Initially, the method estimates nonrigid motion (NME) from the image sequences, using intensity-based image registration. The motion-compensated image sequence is then averaged to obtain a single 'epitome' image with improved signal-to-noise ratio. The lumen is segmented from the epitome image through an intensity joint-histogram classification and a graph-based segmentation. NME was validated by comparing displacements with manual annotations in eleven carotids. The average root-mean-squareerror (RMSE) was 112 73 μm. Segmentation results were validated against manual delineations in the epitome images of two different datasets, respectively containing eleven (RMSE 191 43 μm) and ten (RMSE 351 176 μm) carotids. From the deformation fields, we derived arterial distensibility with values comparable to the literature. The average errors in all experiments were in the inter-observer variability range. To the best of our knowledge, this is the first study exploiting combined BMUS&CEUS images for atherosclerotic carotid lumen segmentation.
IEEE Transactions on Medical Imaging 11/2014; 34(4). DOI:10.1109/TMI.2014.2372784 · 3.39 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Background
To evaluate the influence of image registration on apparent diffusion coefficient (ADC) images obtained from abdominal free-breathing diffusion-weighted MR images (DW-MRIs).MethodsA comprehensive pipeline based on automatic three-dimensional nonrigid image registrations is developed to compensate for misalignments in DW-MRI datasets obtained from five healthy subjects scanned twice. Motion is corrected both within each image and between images in a time series. ADC distributions are compared with and without registration in two abdominal volumes of interest (VOIs). The effects of interpolations and Gaussian blurring as alternative strategies to reduce motion artifacts are also investigated.ResultsAmong the four considered scenarios (no processing, interpolation, blurring and registration), registration yields the best alignment scores. Median ADCs vary according to the chosen scenario: for the considered datasets, ADCs obtained without processing are 30% higher than with registration. Registration improves voxelwise reproducibility at least by a factor of 2 and decreases uncertainty (Fréchet-Cramér-Rao lower bound). Registration provides similar improvements in reproducibility and uncertainty as acquiring four times more data.Conclusion
Patient motion during image acquisition leads to misaligned DW-MRIs and inaccurate ADCs, which can be addressed using automatic registration. J. Magn. Reson. Imaging 2014.
Journal of Magnetic Resonance Imaging 11/2014; 42(2). DOI:10.1002/jmri.24792 · 3.21 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: To assess (1) whether normal and degenerated menisci exhibit different T1GD on delayed gadolinium-enhanced MRI of the meniscus (dGEMRIM), (2) the reproducibility of dGEMRIM and (3) the correlation between meniscus and cartilage T1GD in knee osteoarthritis (OA) patients.
In 17 OA patients who underwent dGEMRIM twice within 7 days, meniscus and cartilage T1GD was calculated. Meniscus pathology was evaluated on conventional MRI. T1GD in normal and degenerated menisci were compared using a Student's t-test. Reproducibility was assessed using ICCs. Pearson's correlation was calculated between meniscus and cartilage T1GD.
A trend towards lower T1GD in degenerated menisci (mean: 402 ms; 95 % CI: 359-444 ms) compared to normal menisci (mean: 448 ms; 95 % CI: 423-473 ms) was observed (p = 0.05). Meniscus T1GD ICCs were 0.85-0.90. The correlation between meniscus and cartilage T1GD was moderate in the lateral (r = 0.52-0.75) and strong in the medial compartment (r = 0.78-0.94).
Our results show that degenerated menisci have a clear trend towards lower T1GD compared to normal menisci. Since these results are highly reproducible, meniscus degeneration may be assessed within one delayed gadolinium-enhanced MRI simultaneously with cartilage. The strong correlation between meniscus and cartilage T1GD suggests concomitant degeneration in both tissues in OA, but also suggests that dGEMRIC may not be regarded entirely as sulphated glycosaminoglycan specific.
• dGEMRIM T1 GD can possibly be used to assess meniscal degeneration; • dGEMRIM yields highly reproducible meniscal T1 GD in early stage osteoarthritic patients; • Concomitant degeneration of cartilage and meniscus tissue occurs in early stage osteoarthritis; • dGEMRIC cannot be regarded as entirely sulphated glycosaminoglycan specific.
European Radiology 05/2014; 24(9). DOI:10.1007/s00330-014-3204-z · 4.01 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: To introduce a semiautomatic algorithm to perform the registration of free-hand B-Mode ultrasound (US) and magnetic resonance imaging (MRI) of the carotid artery.
The authors' approach combines geometrical features and intensity information. The only user interaction consists of placing three seed points in US and MRI. First, the lumen centerlines are used as landmarks for point based registration. Subsequently, in a joint optimization the distance between centerlines and the dissimilarity of the image intensities is minimized. Evaluation is performed in left and right carotids from six healthy volunteers and five patients with atherosclerosis. For the validation, the authors measure the Dice similarity coefficient (DSC) and the mean surface distance (MSD) between carotid lumen segmentations in US and MRI after registration. The effect of several design parameters on the registration accuracy is investigated by an exhaustive search on a training set of five volunteers and three patients. The optimum configuration is validated on the remaining images of one volunteer and two patients.
On the training set, the authors achieve an average DSC of 0.74 and a MSD of 0.66 mm on volunteer data. For the patient data, the authors obtain a DSC of 0.77 and a MSD of 0.69 mm. In the independent set composed of patient and volunteer data, the DSC is 0.69 and the MSD is 0.87 mm. The experiments with different design parameters show that nonrigid registration outperforms rigid registration, and that the combination of intensity and point information is superior to approaches that use intensity or points only.
The proposed method achieves an accurate registration of US and MRI, and may thus enable multimodal analysis of the carotid plaque.
Medical Physics 05/2014; 41(5):052904. DOI:10.1118/1.4870383 · 2.64 Impact Factor
[Show abstract][Hide abstract] ABSTRACT: Atherosclerotic plaque composition can indicate plaque vulnerability. We segment atherosclerotic plaque components from the carotid artery on a combination of in vivo MRI and CT-angiography (CTA) data using supervised voxelwise classification. In contrast to previous studies the ground truth for training is directly obtained from 3D registration with histology for fibrous and lipid-rich necrotic tissue, and with CT for calcification. This registration does, however, not provide accurate voxelwise correspondence. We therefore evaluate three approaches that incorporate uncertainty in the ground truth used for training: I) soft labels are created by Gaussian blurring of the original binary histology segmentations to reduce weights at the boundaries between components, and are weighted by the estimated registration accuracy of the histology and in vivo imaging data (measured by overlap), II) samples are weighted by the local contour distance of the lumen and outer wall between histology and in vivo data, and III) 10% of each class is rejected by Gaussian outlier rejection. Classification was evaluated on the relative volumes (% of tissue type in the vessel wall) for calcified, fibrous and lipid-rich necrotic tissue, using linear discriminant (LDC) and support vector machine (SVM) classification. In addition, the combination of MRI and CTA data was compared to using only one imaging modality. Best results were obtained by LDC and outlier rejection: the volume error per vessel was 0.91.0% for calcification, 12.77.6% for fibrous and 12.18.1% for necrotic tissue, with Spearman rank correlation coefficients of 0.91 (calcification), 0.80 (fibrous) and 0.81 (necrotic). While segmentation using only MRI features yielded low accuracy for calcification, and segmentation using only CTA features yielded low accuracy for necrotic tissue, the combination of features from MRI and CTA gave good results for all studied components.
PLoS ONE 04/2014; 9(4):e94840. DOI:10.1371/journal.pone.0094840 · 3.23 Impact Factor