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

Abstract

Brain morphometry based classification from magnetic resonance (MR) acquisitions has been widely investigated in the diagnosis of Alzheimer's disease (AD) and its prodromal stage, i.e., mild cognitive impairment (MCI). In the literature, a morphometric representation of brain structures is obtained by spatial normalization of each image into a common space (i.e., a pre-defined atlas) via non-linear registration, thus the corresponding regions in different brains can be compared. However, representations generated from one single atlas may not be sufficient to reveal the underlying anatomical differences between the groups of disease-affected patients and normal controls (NC). In this article, we propose a different methodology, namely the multi-atlas based morphometry, which measures morphometric representations of the same image in different spaces of multiple atlases. Representations generated from different atlases can thus provide the complementary information to discriminate different groups, and also reduce the negative impacts from registration errors. Specifically, each studied subject is registered to multiple atlases, where adaptive regional features are extracted. Then, all features from different atlases are jointly selected by a correlation and relevance based scheme, followed by final classification with the support vector machine (SVM). We have evaluated the proposed method on 459 subjects (97 AD, 117 progressive-MCI (p-MCI), 117 stable-MCI (s-MCI), and 128 NC) from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database, and achieved 91.64% for AD/NC classification and 72.41% for p-MCI/s-MCI classification. Our results clearly demonstrate that the proposed multi-atlas based method can significantly outperform the previous single-atlas based methods. Hum Brain Mapp, 2014. © 2014 Wiley Periodicals, Inc.

No full-text available

Request Full-text Paper PDF

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

... and dimensionality reduction algorithms were quite similar; (3) the GMV and ReHo features in the prefrontal and temporal gyri made the greatest contributions in both classifications; and (4) the ensemble learning method substantially improved classification performance. Generally, the selection of the brain atlas may result in striking differences in performance of the classification of psychiatric diseases (Koikkalainen et al., 2011;Min et al., 2014;Asim et al., 2018;Kalmady et al., 2019). Kalmady et al. (2019) used 14 brain atlases for discriminative analyses of SZ patients and found that the accuracies of the classifications varied significantly across different brain atlases. ...
... Kalmady et al. (2019) used 14 brain atlases for discriminative analyses of SZ patients and found that the accuracies of the classifications varied significantly across different brain atlases. Similarly, a number of discriminative analyses have also been performed with patients with Alzheimer's disease (AD) based on multiple brain atlases (Koikkalainen et al., 2011;Min et al., 2014;Asim et al., 2018), in which the features based on all atlases were used to establish the integral model and achieved the best classification performance (Koikkalainen et al., 2011;Min et al., 2014;Asim et al., 2018). Consistent with previous studies, our results also showed discrepancies in the classification performance with different brain atlases. ...
... Kalmady et al. (2019) used 14 brain atlases for discriminative analyses of SZ patients and found that the accuracies of the classifications varied significantly across different brain atlases. Similarly, a number of discriminative analyses have also been performed with patients with Alzheimer's disease (AD) based on multiple brain atlases (Koikkalainen et al., 2011;Min et al., 2014;Asim et al., 2018), in which the features based on all atlases were used to establish the integral model and achieved the best classification performance (Koikkalainen et al., 2011;Min et al., 2014;Asim et al., 2018). Consistent with previous studies, our results also showed discrepancies in the classification performance with different brain atlases. ...
Full-text available
Article
Recently, machine learning techniques have been widely applied in discriminative studies of schizophrenia (SZ) patients with multimodal magnetic resonance imaging (MRI); however, the effects of brain atlases and machine learning methods remain largely unknown. In this study, we collected MRI data for 61 first-episode SZ patients (FESZ), 79 chronic SZ patients (CSZ) and 205 normal controls (NC) and calculated 4 MRI measurements, including regional gray matter volume (GMV), regional homogeneity (ReHo), amplitude of low-frequency fluctuation and degree centrality. We systematically analyzed the performance of two classifications (SZ vs NC; FESZ vs CSZ) based on the combinations of three brain atlases, five classifiers, two cross validation methods and 3 dimensionality reduction algorithms. Our results showed that the groupwise whole-brain atlas with 268 ROIs outperformed the other two brain atlases. In addition, the leave-one-out cross validation was the best cross validation method to select the best hyperparameter set, but the classification performances by different classifiers and dimensionality reduction algorithms were quite similar. Importantly, the contributions of input features to both classifications were higher with the GMV and ReHo features of brain regions in the prefrontal and temporal gyri. Furthermore, an ensemble learning method was performed to establish an integrated model, in which classification performance was improved. Taken together, these findings indicated the effects of these factors in constructing effective classifiers for psychiatric diseases and showed that the integrated model has the potential to improve the clinical diagnosis and treatment evaluation of SZ.
... A potential issue of using a single atlas for all subjects is that it assumes a highly consistent functional border of every brain region across all subjects. However, such an assumption may not hold in practice, due to not only imperfect spatial registration [28], [29] but also mismatched anatomical-functional correspondence [30]. In fact, evidence has suggested that different subjects could have different functional parcellation, with the varied functional border of brain regions [30]. ...
... For instance, Koikkalainen et al. [42] developed a multiatlas-based strategy to reduce the MRI registration bias for AD classification and obtained increased accuracy in comparison with a single-atlas-based method. Min et al. [29] proposed to generate multiple feature representations of the anatomical structures based on multiple atlases. The derived features were then aggregated to form a comprehensive representation for enhancing the separability between AD and healthy subjects. ...
... So far, many multiatlas methods have been developed for improved brain disease diagnosis by exploiting multiple anatomical templates [28], [29], [42], [43]. However, all these existing methods were specifically designed for structural MRI analysis. ...
Article
Functional connectivity (FC) networks built from resting-state functional magnetic resonance imaging (rs-fMRI) has shown promising results for the diagnosis of Alzheimer's disease and its prodromal stage, that is, mild cognitive impairment (MCI). FC is usually estimated as a temporal correlation of regional mean rs-fMRI signals between any pair of brain regions, and these regions are traditionally parcellated with a particular brain atlas. Most existing studies have adopted a predefined brain atlas for all subjects. However, the constructed FC networks inevitably ignore the potentially important subject-specific information, particularly, the subject-specific brain parcellation. Similar to the drawback of the "single view" (versus the "multiview" learning) in medical image-based classification, FC networks constructed based on a single atlas may not be sufficient to reveal the underlying complicated differences between normal controls and disease-affected patients due to the potential bias from that particular atlas. In this study, we propose a multiview feature learning method with multiatlas-based FC networks to improve MCI diagnosis. Specifically, a three-step transformation is implemented to generate multiple individually specified atlases from the standard automated anatomical labeling template, from which a set of atlas exemplars is selected. Multiple FC networks are constructed based on these preselected atlas exemplars, providing multiple views of the FC network-based feature representations for each subject. We then devise a multitask learning algorithm for joint feature selection from the constructed multiple FC networks. The selected features are jointly fed into a support vector machine classifier for multiatlas-based MCI diagnosis. Extensive experimental comparisons are carried out between the proposed method and other competing approaches, including the traditional single-atlas-based method. The results indicate that our method significantly improves the MCI classification, demonstrating its promise in the brain connectome-based individualized diagnosis of brain diseases.
... In contrast, multi-template based methods are more promising to discover disease status and compare group difference ( Liu et al., 2016 b). It is suggested in the previous studies ( Jin et al., 2015 ;Liu et al., 2016 a;Min et al., 2014 ) that learning with multiple templates can boost diagnosis accuracy. For example, Min et al. ( Min et al., 2014 ) utilized concate-nated multi-template based features of each subject and achieved promising AD classification results. ...
... It is suggested in the previous studies ( Jin et al., 2015 ;Liu et al., 2016 a;Min et al., 2014 ) that learning with multiple templates can boost diagnosis accuracy. For example, Min et al. ( Min et al., 2014 ) utilized concate-nated multi-template based features of each subject and achieved promising AD classification results. Multiple templates not only represent the brain information in a comprehensive way, but also capture the disease-related discriminative information ( Liu et al., 2016 b). ...
... Multiple templates not only represent the brain information in a comprehensive way, but also capture the disease-related discriminative information ( Liu et al., 2016 b). Also, multi-template based methods can extract multiple feature sets of a subject derived from different templates ( Jin et al., 2015 ;Liu et al., 2016 a;Min et al., 2014 ), which can effectively reduce the negative impacts of registration errors and provide distinct yet complementary information to identify different disease status. It thus leads to more promising identification performance. ...
... For all these reasons, anatomical MRI has been the most widely used modality for developing ML approaches for AD classification and prediction. Based on whole-brain data, AD vs CN classification is usually highly accurate, with accuracies ranging from 80% to 95 % (Aguilar et al., 2013;Bron et al., 2015;Cuingnet et al., 2011;Farhan, Fahiem, and Tauseef, 2014;Gerardin et al., 2009;Klöppel et al., 2008b;Magnin et al., 2009;Min et al., 2014;Minhas et al., 2018;Tong et al., 2014;Vemuri et al., 2008;Zhou et al., 2014;Coupé et al., 2012a). On the other hand, the results for predicting conversion of MCI patients are not good, with accuracies from 60% to 80% for the classification of sMCI vs pMCI (Aguilar et al., 2013;Chupin et al., 2009;Cuingnet et al., 2011;Min et al., 2014;Tong et al., 2014;Adaszewski et al., 2013;Plant et al., 2010;Costafreda et al., 2011;Sørensen et al., 2016). ...
... Based on whole-brain data, AD vs CN classification is usually highly accurate, with accuracies ranging from 80% to 95 % (Aguilar et al., 2013;Bron et al., 2015;Cuingnet et al., 2011;Farhan, Fahiem, and Tauseef, 2014;Gerardin et al., 2009;Klöppel et al., 2008b;Magnin et al., 2009;Min et al., 2014;Minhas et al., 2018;Tong et al., 2014;Vemuri et al., 2008;Zhou et al., 2014;Coupé et al., 2012a). On the other hand, the results for predicting conversion of MCI patients are not good, with accuracies from 60% to 80% for the classification of sMCI vs pMCI (Aguilar et al., 2013;Chupin et al., 2009;Cuingnet et al., 2011;Min et al., 2014;Tong et al., 2014;Adaszewski et al., 2013;Plant et al., 2010;Costafreda et al., 2011;Sørensen et al., 2016). ...
... For example, the boundaries of the atlas regions may not correspond to the boundaries of the disease alterations. To overcome this drawback, some authors have proposed disease specific parcellations (Min et al., 2014;Fan et al., 2007). ...
Thesis
Alzheimer’s disease (AD) is the first cause of dementia worldwide, affectingover 20 million people. Its diagnosis at an early stage is essential to ensure aproper care of patients, and to develop and test novel treatments. AD is a complexdisease that has to be characterized by the use of different measurements: cognitiveand clinical tests, neuroimaging including magnetic resonance imaging (MRI)and positron emission tomography (PET), genotyping, etc. There is an interest inexploring the discriminative and predictive capabilities of these diverse markers,which reflect different aspects of the disease and potentially carry complementaryinformation, from an early stage of the disease.The objective of this PhD thesis was thus to assess the potential and to integratemultiple modalities using machine learning methods, in order to automaticallyclassify patients with AD and predict the development of the disease from the earlieststages. More specifically, we aimed to make progress toward the translationof such approaches toward clinical practice.The thesis comprises three main studies. The first one tackles the differentialdiagnosis between different forms of dementia from MRI data. This study was performedusing clinical routine data, thereby providing a more realistic evaluationscenario. The second one proposes a new framework for reproducible evaluationof AD classification algorithms from MRI and PET data. Indeed, while numerousapproaches have been proposed for AD classification in the literature, theyare difficult to compare and to reproduce. The third part is devoted to the predictionof progression to AD in patients with mild cognitive impairment throughthe integration of multimodal data, including MRI, PET, clinical/cognitive evaluationsand genotyping. In particular, we systematically assessed the added valueof neuroimaging over clinical/cognitive data only. Since neuroimaging is moreexpensive and less widely available, this is important to justify its use as input ofclassification algorithms.
... In "fixed atlas based methods," 4,18 brains are parcellated into anatomical regions based on fixed pre-defined atlases, and features are extracted from those regions. However, in "adaptive atlas based methods," 19,20 adaptive regions are calculated based on the subjects involved in one particular study, and features are calculated from the adaptively calculated regions. ...
... In addition, two methods from adaptive atlas-based methods were also chosen for comparison. First, Min et al. derived multiple atlases from the non-overlapping clusters of subjects, 19 obtained using affinity propagation. 38 They registered subjects to the atlases and adaptively calculated a set of ROIs and volumetric features in each atlas space. ...
... Subsequently, Liu et al. argued that the features extracted from K sets of adaptive ROIs are different representations of the same subject, 20 and should not be concatenated, as in a previous study. 19 To resolve this, Liu et al. registered subjects to different selected atlases and extracted features from adaptive regions of each atlas-registered image, viewing that image as the main source, and all other atlas registeredimages as adjunctive sources. 20 SVM was separately trained on features extracted from each set and the results of multiple sets were combined using majority voting. ...
Full-text available
Article
The use of biomarkers for early detection of Alzheimer's disease (AD) improves the accuracy of imaging-based prediction of AD and its prodromal stage that is mild cognitive impairment (MCI). Brain parcellation-based computer-aided methods for detecting AD and MCI segregate the brain in different anatomical regions and use their features to predict AD and MCI. Brain parcellation generally is carried out based on existing anatomical atlas templates, which vary in the boundaries and number of anatomical regions. This works considers dividing the brain based on different atlases and combining the features extracted from these anatomical parcellations for a more holistic and robust representation. We collected data from the ADNI database and divided brains based on two well-known atlases: LONI Probabilistic Brain Atlas (LPBA40) and Automated Anatomical Labeling (AAL). We used baselines images of structural magnetic resonance imaging (MRI) and ¹⁸F-fluorodeoxyglucose positron emission tomography (FDG-PET) to calculate average gray-matter density and average relative cerebral metabolic rate for glucose in each region. Later, we classified AD, MCI and cognitively normal (CN) subjects using the individual features extracted from each atlas template and the combined features of both atlases. We reduced the dimensionality of individual and combined features using principal component analysis, and used support vector machines for classification. We also ranked features mostly involved in classification to determine the importance of brain regions for accurately classifying the subjects. Results demonstrated that features calculated from multiple atlases lead to improved performance compared to those extracted from one atlas only.
... Liu et al. (2014a) developed a multitask feature selection method to preserve inter-modality relationship by imposing a constraint and employed support vector machine (SVM) to combine the significant features for AD diagnosis. Min et al. (2014) registered each subject with multiple atlases and calculated the correlation among them to select the relevant features then used the SVM for classification. Zhang et al. (2012) developed a multimodal multi-task learning that selected the significant features from each modality. ...
... Then, N3 algorithm was applied to correct the bias field of the processed AC-PC images. Afterward, we perform the brain extraction using the skull-stripping technique in Min et al. (2014) followed by removing the cerebellum. Furthermore, we register the skullstripped image with Montreal Neurological Institute template (Lancaster et al., 2000). ...
Full-text available
Article
Alzheimer’s disease (AD) is an age-related disease that affects a large proportion of the elderly. Currently, the neuroimaging techniques [e.g., magnetic resonance imaging (MRI) and positron emission tomography (PET)] are promising modalities for AD diagnosis. Since not all brain regions are affected by AD, a common technique is to study some region-of-interests (ROIs) that are believed to be closely related to AD. Conventional methods used ROIs, identified by the handcrafted features through Automated Anatomical Labeling (AAL) atlas rather than utilizing the original images which may induce missing informative features. In addition, they learned their framework based on the discriminative patches instead of full images for AD diagnosis in multistage learning scheme. In this paper, we integrate the original image features from MRI and PET with their ROIs features in one learning process. Furthermore, we use the ROIs features for forcing the network to focus on the regions that is highly related to AD and hence, the performance of the AD diagnosis can be improved. Specifically, we first obtain the ROIs features from the AAL, then we register every ROI with its corresponding region of the original image to get a synthetic image for each modality of every subject. Then, we employ the convolutional auto-encoder network for learning the synthetic image features and the convolutional neural network (CNN) for learning the original image features. Meanwhile, we concatenate the features from both networks after each convolution layer. Finally, the highly learned features from the MRI and PET are concatenated for brain disease classification. Experiments are carried out on the ADNI datasets including ADNI-1 and ADNI-2 to evaluate our method performance. Our method demonstrates a higher performance in brain disease classification than the recent studies.
... In addition to predefined region-based methods, density maps and cortical surface are other main feature extraction methods that can be used in sMRI for AD/NC classification. In studies using the density map-based method, researchers used either whole-density maps [22][23][24][25][26][27] or reduced-density maps as features, which are features that have been reduced using supervised or unsupervised feature-reduction methods [28][29][30] or features that have been extracted from predefined atlases [31] or adaptive regions to reduce dimensionality [32][33][34]. For studies using the surface-based method, the authors used supervised/ unsupervised feature reduction-based methods [35,36] or atlas-based methods to reduce vertices as features [37][38][39][40][41][42]. ...
... Salvatore et al. [29] implemented PCA coupled with a Fisher discriminant ratio criterion to MR images (whole-brain, GM, and white matter) and obtained an accuracy of 66%. Min et al. [33] extracted adaptive regional features and yielded an accuracy of 72.41%. Liu et al. [34] proposed a novel view-centralized multiatlas classification method and achieved an accuracy of 78.88% for cMCI/ncMCI classification. ...
Full-text available
Article
The use of neuroimaging examinations is crucial in Alzheimer’s disease (AD), in both research and clinical settings. Over the years, magnetic resonance imaging (MRI)–based computer‐aided diagnosis has been shown to be helpful for early screening and predicting cognitive decline. Meanwhile, an increasing number of studies have adopted machine learning for the classification of AD, with promising results. In this review article, we focus on computerized MRI‐based biomarkers of AD by reviewing representative studies that used computerized techniques to identify AD patients and predict cognitive progression. We categorized these studies based on the following applications: (1) identifying AD from normal control; (2) identifying AD from other dementia types, including vascular dementia, dementia with Lewy bodies, and frontotemporal dementia; and (3) predicting conversion from NC to mild cognitive impairment (MCI) and from MCI to AD. This systematic review could act as a state‐of‐the‐art overview of this emerging field as well as a basis for designing future studies.
... In order to suggest a multi Atlas-based morphometry technique that calculates the morphometric representations of the same picture in different spaces of multiple atlases, Min et al. [19] tested a sample of 459 subjects consisting of 97 AD, 117 progressive MCI (p-MCI), 117 stable MCI (s-MCI), and 128 NC patients from ADNI databases. Their data are recorded in several atlases, in which adaptive regional characteristics are extracted. ...
... views). The authors argued that the features isolated from K adaptive ROI sets are representations of the same subject separately, so they should not be related, as previously stated in Min et al. [19]. They reported the subjects to various selected atlases as a solution. ...
Full-text available
Article
Neuroimaging is a description, whether in two-dimensions (2D) or three-dimensions (3D), of the structure and functions of the brain. Neuroimaging provides a valuable diagnostic tool, in which a limited approach is used to create images of the focal sensory system by medicine professionals. For the clinical diagnosis of patients with Alzheimer's Disease (AD) or Mild Cognitive Impairs (MCI), the accurate identification of patients from normal control persons (NCs) is critical. Recently, numerous researches have been undertaken on the identification of AD based on neuroimaging data, including images with radiographs and algorithms for master learning. In the previous decade, these techniques were also used slowly to differentiate AD and MCI symptoms from structure classification methods. This review focuses on neuroimaging studies conducted to detect and classify AD, through a survey based on Google Scholar content. We explore the challenges of this field and evaluate the performance of these studies along with their negative aspects.
... R. Min, G. Wu, J. Cheng, Q. Wang, and D. Shen proposed a multi-atlas based representations for Alzheimer's disease diagnosis using Data-driven ROI GM and SVM classifier [19]. This method provides an accuracy rate of 91%. ...
... SVM and Wavelet Transform [14] 94% Shape-constrained regression-forest algorithm and SVM [15] 83.7% Grading Biomarker using Sparse representation techniques and SVM [16] 92% Voxel-wise GM,LDS [17] 83% Data-driven ROI GM, SVM ensemble [18] 92.51% Data-driven ROI GM,SVM [19] 91% ROI-wise cortical thickness, LDA [20] 84.50% Voxel-wise GM, RVR [21] 84.6% Ensemble Random forests [22] 89.0% Voxel-Stand-D GM,SVM [23] 88.58% ROI GM,SVM [24] 86.20% Tensor-base Morphometry, Linear regression [25] 86% 75% 80% 85% 90% 95% 100% Accuracy Measurements ...
... Although we explored a relatively long 5-year follow-up, it is likely that a nonnegligible proportion of our stable MCI over 5 years is still affected by AD and presented evident atrophies at baseline, as suggested by the fact that we obtained a significant number of false positives. In this context, (Liu et al., 2013) 97p/93 s CV 3 72% Eskildsen et al. (2013) (Eskildsen et al., 2013) 128p/227 s CV Variable (Mean = 1.5 years) 68% Min et al. (2014) (Min et al., 2014) 98AD/128NC CV (117p/117 s) Not reported 67% Liu et al. (2015) (Liu et al., 2015) 117p/117 s CV Not reported 81% Tang et al. (2015) (Tang et al., 2015) 175AD/210NC CV (135p/87 s)) 3 74% Chincarini et al. (2011) (Chincarini et al., 2011) 144AD/189NC Independent set (136p/166 s) 2 74% Wee et al. (2013) (Wee et al., 2013) 45p/56 s Repeated hold-out (44p/55 s) 3 84% Sorensen et al. (2016) (Sorensen et al., 2016) 101AD A. Moscoso, et al. NeuroImage: Clinical 23 (2019) 101837 and given the lack of data with long follow-ups, the usually abandoned strategy of using NC vs. AD for training might be more adequate than what was expected. ...
... Although we explored a relatively long 5-year follow-up, it is likely that a nonnegligible proportion of our stable MCI over 5 years is still affected by AD and presented evident atrophies at baseline, as suggested by the fact that we obtained a significant number of false positives. In this context, (Liu et al., 2013) 97p/93 s CV 3 72% Eskildsen et al. (2013) (Eskildsen et al., 2013) 128p/227 s CV Variable (Mean = 1.5 years) 68% Min et al. (2014) (Min et al., 2014) 98AD/128NC CV (117p/117 s) Not reported 67% Liu et al. (2015) (Liu et al., 2015) 117p/117 s CV Not reported 81% Tang et al. (2015) (Tang et al., 2015) 175AD/210NC CV (135p/87 s)) 3 74% Chincarini et al. (2011) (Chincarini et al., 2011) 144AD/189NC Independent set (136p/166 s) 2 74% Wee et al. (2013) (Wee et al., 2013) 45p/56 s Repeated hold-out (44p/55 s) 3 84% Sorensen et al. (2016) (Sorensen et al., 2016) 101AD A. Moscoso, et al. NeuroImage: Clinical 23 (2019) 101837 and given the lack of data with long follow-ups, the usually abandoned strategy of using NC vs. AD for training might be more adequate than what was expected. ...
Full-text available
Article
Magnetic resonance imaging (MRI) volumetric measures have become a standard tool for the detection of incipient Alzheimer's Disease (AD) dementia in mild cognitive impairment (MCI). Focused on providing an earlier and more accurate diagnosis, sophisticated MRI machine learning algorithms have been developed over the recent years, most of them learning their non-disease patterns from MCI that remained stable over 2–3 years. In this work, we analyzed whether these stable MCI over short-term periods are actually appropriate training examples of non-disease patterns. To this aim, we compared the diagnosis of MCI patients at 2 and 5 years of follow-up, and investigated its impact on predictive performance of baseline volumetric MRI measures primarily involved in AD, i.e., hippocampal and entorhinal cortex volumes. Predictive power was evaluated in terms of the area under the ROC curve (AUC), sensitivity, and specificity in a trial sample of 248 MCI patients followed-up over 5 years. We further compared the sensitivity in those MCI that converted before 2 years and those that converted after 2 years. Our results indicate that 23% of the stable MCI at 2 years progressed in the next three years and that MRI volumetric measures are good predictors of conversion to AD dementia even at the mid-term, showing a better specificity and AUC as follow-up time increases. The combination of hippocampus and entorhinal cortex yielded an AUC that was significantly higher for the 5-year follow-up (AUC = 73% at 2 years vs. AUC = 84% at 5 years), as well as for specificity (56% vs. 71%). Sensitivity showed a non-significant slight decrease (81% vs. 78%). Remarkably, the performance of this model was comparable to machine learning models at the same follow-up times. MRI correctly identified most of the patients that converted after 2 years (with sensitivy >60%), and these patients showed a similar degree of abnormalities to those that converted before 2 years. This implies that most of the MCI patients that remained stable over short periods and subsequently progressed to AD dementia had evident atrophies at baseline. Therefore, machine learning models that use these patients to learn non-disease patterns are including an important fraction of patients with evident pathological changes related to the disease, something that might result in reduced performance and lack of biological interpretability.
... Multi-atlas based methods have been recently used for image segmentation [19], image registration [1], and disease classification [22,28,35]. Compared with conventional single atlas based methods, multi-atlas based methods adopt multiple predefined atlases, and thus can obtain more complementary information. ...
... Experimental results showed that better results can be obtained by using multiple registration in the diagnosis of Alzheimer's disease. To obtain a rich feature representation for Alzheimer's disease classification, Min et al. [35] proposed to measure brain gray matter morphometry via multiple atlases as feature representation. Unlike the work of Koikkalainen et al. [22] which registers all selected atlases to a common space by adopting deformable registration method, they proposed to register subjects from different groups to different atlases in order to consider different effective information provided by different atlases. ...
Full-text available
Article
Schizophrenia (SZ) is a complex neuropsychiatric disorder that seriously affects the daily life of patients. Therefore, accurate diagnosis of SZ is essential for patient care. Several T1-weighted magnetic resonance imaging (MRI) and diffusion tensor imaging (DTI) markers (e.g., cortical thickness (CT), mean diffusivity (MD)) for SZ have been identified by using some existing brain atlases, and have been used successfully to discriminate patients with SZ from healthy controls (HCs). Currently, these markers have mainly been used separately. Thus, the full potential of T1-weighted MRI images and DTI images for SZ diagnosis might not yet have been used comprehensively. Furthermore, the extraction of these markers based on single brain atlas might not yet be able to use the full potential of these images. Therefore, in this study, we propose a multi-modality multi-atlas feature representation and a multi-kernel learning method (MMM) to perform SZ/HC classification. Firstly, we extract 8 feature sets from T1-weighted MRI images and DTI images via 4 existing brain atlases and 4 markers. Then, a two-step feature selection method is proposed to select the most discriminative features of each feature set for SZ/HC classification. Finally, a multiple feature sets based multi-kernel SVM learning method (MFMK-SVM) is proposed to combine all feature sets for SZ/HC classification. Experimental results show that our proposed method achieves an accuracy of 91.28%, a sensitivity of 90.85%, a specificity of 92.17% and an AUC of 0.9485 for SZ/HC classification. Experimental results illustrate that our proposed classification method is efficient and promising for clinical diagnosis of SZ.
... In this method, subjects are registered to multiple atlases, and adaptive ROIs and corresponding regional volumetric measures are calculated in each atlas space to overcome the inherent bias associated with one atlas. For example, Min et al. derived multiple atlases from the non-overlapping clusters of subjects (Min et al., 2014), obtained using affinity propagation (Frey, Dueck, 2007). They registered subjects to the atlases and adaptively calculated a set of ROIs and volumetric features in each atlas space. ...
... The top-most K discriminating features calculated from each atlas space were combined for SVM-based classification. Subsequently, Liu et al. argued that the features extracted from K sets of adaptive ROIs are different representations of the same subject , and should not be concatenated, as in a previous study (Min et al., 2014). To resolve this, Liu et al. registered subjects to different selected atlases and extracted features from adaptive regions of each atlas-registered image, viewing that image as the main source, and all other atlas registered-images as adjunctive sources . ...
Full-text available
Article
Neuroimaging has made it possible to measure pathological brain changes associated with Alzheimer's disease (AD) in vivo. Over the past decade, these measures have been increasingly integrated into imaging signatures of AD by means of classification frameworks, offering promising tools for individualized diagnosis and prognosis. We reviewed neuroimaging-based studies for AD classification and mild cognitive impairment, selected after online database searches in Google Scholar and PubMed (January, 1985 to June, 2016). We categorized these studies based on the following neuroimaging modalities (and sub-categorized based on features extracted as a post-processing step from these modalities): i) structural magnetic resonance imaging [MRI] (tissue density, cortical surface, and hippocampal measurements), ii) functional MRI (functional coherence of different brain regions, and the strength of the functional connectivity), iii) diffusion tensor imaging (patterns along the white matter fibers), iv) fluorodeoxyglucose positron emission tomography (metabolic rate of cerebral glucose), and v) amyloid-PET (amyloid burden). The studies reviewed indicate that the classification frameworks formulated on the basis of these features show promise for individualized diagnosis and prediction of clinical progression. Finally, we provided a detailed account of AD classification challenges and address some future research directions.
... Our focus on the hippocampus was related to the research question of the relationship between variations in insulin resistance, memory and eating behavior in adolescents (Mucellini et al., 2017). MAGeT-Brain minimizes the number of input atlases needed by creating a template library from a sample of the subject images, differently from the voxel-based morphometry (VMB), which is performed by first spatially normalizing all subjects into a common atlas space (de Flores et al., 2015;Min et al., 2014). Using MAGeT-Brain, volumes for the hippocampus and its subfields (total hippocampal volume, CA4 and DG, CA2 and CA3, CA1, subiculum and stratum radiatum, lacunosum and moleculare, right and left) were extracted for each participant for statistical comparisons (please see statistical analysis section below). ...
Article
While classically linked to memory, the hippocampus is also a feeding behavior modulator due to its multiple interconnected pathways with other brain regions and expression of receptor for metabolic hormones. Here we tested whether variations in insulin sensitivity would be correlated with differential brain activation following exposure to palatable food cues, as well as with variations in implicit food memory in a cohort of healthy adolescents, some of whom were born small for gestational age (SGA). Homeostatic Model Assessment of Insulin Resistance (HOMA-IR) was positively correlated with activation in the cuneus, and negatively correlated with activation in the middle frontal lobe, superior frontal gyrus and precuneus when presented with palatable food images versus non-food images in healthy adolescents. Additionally, HOMA-IR and insulinemia were higher in participants with impaired food memory. SGA individuals had higher snack caloric density and greater chance for impaired food memory. There was also an interaction between the HOMA-IR and birth weight ratio influencing external eating behavior. We suggest that diminished insulin sensitivity correlates with activation in visual attention areas and inactivation in inhibitory control areas in healthy adolescents. Insulin resistance also associated with less consistency in implicit memory for a consumed meal, which may suggest lower ability to establish a dietary pattern, and can contribute to obesity. Differences in feeding behavior in SGA individuals were associated with insulin sensitivity and hippocampal alterations, suggesting that cognition and hormonal regulation are important components involved in food intake modifications throughout life.
... The Harris feature detector is also used for corner detection. 7,8 It discovers Harris corners using the Harris affine detection on nine pre-selected scales and two additional scales surrounding the most populated one and culling weak points using a measure derived from the Hessian determinant. ...
... [36], including motion correction, non-uniform intensity normalization, talairach transform computation, skull removal, volumetric segmentation, cortical surface reconstruction and so on. After this standard preprocessing procedure, we can obtain the gray matter (GM) map, which lies between the gray-white interface and the pial surface, and has been widely used to investigate AD/MCI in the literatures [37,38]. Since EMCI is accompanied by brain atrophy, we suspect that the gray matter volume (GMV) is also accompanied by a decrease. ...
Full-text available
Article
Background The identification of early mild cognitive impairment (EMCI), which is an early stage of Alzheimer’s disease (AD) and is associated with brain structural and functional changes, is still a challenging task. Recent studies show great promises for improving the performance of EMCI identification by combining multiple structural and functional features, such as grey matter volume and shortest path length. However, extracting which features and how to combine multiple features to improve the performance of EMCI identification have always been a challenging problem. To address this problem, in this study we propose a new EMCI identification framework using multi-modal data and graph convolutional networks (GCNs). Firstly, we extract grey matter volume and shortest path length of each brain region based on automated anatomical labeling (AAL) atlas as feature representation from T1w MRI and rs-fMRI data of each subject, respectively. Then, in order to obtain features that are more helpful in identifying EMCI, a common multi-task feature selection method is applied. Afterwards, we construct a non-fully labelled subject graph using imaging and non-imaging phenotypic measures of each subject. Finally, a GCN model is adopted to perform the EMCI identification task. Results Our proposed EMCI identification method is evaluated on 210 subjects, including 105 subjects with EMCI and 105 normal controls (NCs), with both T1w MRI and rs-fMRI data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database. Experimental results show that our proposed framework achieves an accuracy of 84.1% and an area under the receiver operating characteristic (ROC) curve (AUC) of 0.856 for EMCI/NC classification. In addition, by comparison, the accuracy and AUC values of our proposed framework are better than those of some existing methods in EMCI identification. Conclusion Our proposed EMCI identification framework is effective and promising for automatic diagnosis of EMCI in clinical practice.
... This feature extraction technique counts the number of occurrences of the orientation of gradients in the localized areas of an image 8 . Table 1 shows the number of feature points detected and matched using the combined feature extraction method. ...
... Classifiers can tackle the problem of AD diagnosis automatically, among of them, SVM has been widely and successfully used, not only for FDG-PET data [53,55,68,71], but also for other modalities, such as MRI [103][104][105], fMRI [106,107], DTI [108,109], or multi-modality [39,61,75,76]. In recent studies on FDG-PET, give a better performance. ...
Thesis
Alzheimer's disease (AD) is becoming the dominant type of neurodegenerative brain disease in elderly people, which is incurable and irreversible for now. It is expected to diagnose its early stage, Mild Cognitive Impairment (MCI), then interventions can be applied to delay the onset. Fluorodeoxyglucose positron emission tomography (FDG-PET) is considered as a significant and effective modality to diagnose AD and the corresponding early phase since it can capture metabolic changes in the brain thereby indicating abnormal regions. Therefore, this thesis is devoted to identify AD from Normal Control (NC) and predict MCI conversion under FDG-PET modality. For this purpose, three independent novel methods are proposed. The first method focuses on developing connectivities among anatomical regions involved in FDG-PET images which are rarely addressed in previous methods. Such connectivities are represented by either similarities or graph measures among regions. Then combined with each region's properties, these features are fed into a designed ensemble classification framework to tackle problems of AD diagnosis and MCI conversion prediction. The second method investigates features to characterize FDG-PET images from the view of spatial gradients, which can link the commonly used features, voxel-wise and region-wise features. The spatial gradient is quantified by a 2D histogram of orientation and expressed in a multiscale manner. The results are given by integrating different scales of spatial gradients within different regions. The third method applies Convolutional Neural Network (CNN) techniques to three views of FDG-PET data, thereby designing the main multiview CNN architecture. Such an architecture can facilitate convolutional operations, from 3D to 2D, and meanwhile consider spatial relations, which is benefited from a novel mapping layer with cuboid convolution kernels. Then three views are combined and make a decision jointly. Experiments conducted on public dataset show that the three proposed methods can achieve significant performance and moreover, outperform most state-of-the-art approaches.
... Elles représentent dans notre cas les différentes métriques d'histogrammes des 5 biomarqueurs :Épaisseur corticale, courbure corticale interne au niveau des gyri, courbure corticale interne au niveau des sillons, courbure corticale externe au niveau des gyri et courbure corticale externe au niveau des sillons. (2015); Min et al. (2014) ont un horizon de 2 ans alors que celui de notreétude est de 3 ans, ce qui complique la prédiction. Ensuite, la différence peut aussi provenir du fait que Vanquin (2015) ait utilisé une méthode de cross-validation pour le paramétrage de ces SVM contrairement a nous qui avons utilisé une grille de recherche. ...
Thesis
Dans le domaine médical, au cours des deux dernières décennies, un nombre croissant de méthodes d'analyse d'images quantitatives ont été développées dont l'analyse par régions d'intérêt, l'analyse voxel à voxel et l'analyse d'histogrammes. Cette dernière est largement utilisée dans la cadre de la recherche sur la sclérose en plaques afin de quantifier les changements pathologiques diffus particulièrement présents dans cette maladie. Un inconvénient de cette approche est que l'ensemble des informations incluses dans l'histogramme n'est pas exploité ; seules des mesures arbitraires sont choisies pour décrire l'histogramme ; incluant la moyenne, la médiane, les centiles... Ainsi dans un premier lieu, nous avons proposé d'intégrer dans un classifieur toute l'information incluse dans l'histogramme et non pas seulement quelques descripteurs locaux, dans le but d'améliorer les performances de classification des populations de sclérose en plaques (groupes dans un essai thérapeutique et in fine groupes de pronostics différents). Par la suite, étant donné que l'histogramme est une estimation trop simpliste d'une distribution de probabilité, nous présentons l'une des applications possibles de la géométrie de l'information sur les distributions de probabilité et démontrons l'intérêt de l'utilisation de la géométrie non-euclidienne dans le contexte de la classification des populations de la maladie d'Alzheimer. Nous avons notamment fait l'analogie avec le domaine de l'aéronautique, plus précisément dans l'étude des retards aéroportuaires. En effet, l'analyse actuellement réalisée se situe au niveau macroscopique et fournit un indicateur de retard moyen, sans tenir compte des mécanismes intermédiaires pouvant conduire au retard final. Ainsi, dans le cadre de la classification des retards aéroportuaires, tout comme dans les applications médicales, nous avons remplacé l'indicateur moyen par un modèle statistique paramétrique plus complet : les distributions de probabilité.
... Comparison between the proposed classification and previous results.individuals, in other words, pMCI24, pMCI12, and also pMCI36, and then, they estimated MRI biomarker enactment in every set disjointedly. As a fact, in 2013, Casanova et al.[48] used 188 of NC subjects with 171 of AD, 153 pMCI, and 182 sMCI patients, and they could achieve 62% for AD vs. NC classification, while Min et al. in[49] by using datadriven ROI could obtain 73% for pMCI vs. sMCI by employing 117 pMCI and 117 sMCI cases. Tong and Gao in[50] succeeded to obtain 76% for AD vs. NC classification. ...
Full-text available
Article
Magnetic resonance (MR) imaging is a widely used imaging modality for detection of brain anatomical variations caused by brain diseases such as Alzheimer's disease (AD) and mild cognitive impairment (MCI). AD considered as an irreversible neurodegenerative disorder with progressive memory impairment moreover cognitive functions, while MCI would be considered as a transitional phase amongst age-related cognitive weakening. Numerous machine learning approaches have been examined, aiming at AD computer-aided diagnosis through employing MR image analysis. Conversely, MR brain image changes could be caused by different effects such as aging and dementia. It is still a challenging difficulty to extract the relevant imaging features and classify the subjects of different groups. This paper would propose an automatic classification technique based on feature decomposition and kernel discriminant analysis (KDA) for classifications of progressive MCI (pMCI) vs. normal control (NC), AD vs. NC, and pMCI vs. stable MCI (sMCI). Feature decomposition would be based on dictionary learning, which is used for separation of class-specific components from the non-class-specific components in the features, while KDA would be applied for mapping original nonlinearly separable feature space to the separable features that are linear. The proposed technique would be evaluated by employing T1-weighted MR brain images from 830 subjects comprising 198 AD patients, 167 pMCI, 236 sMCI, and 229 NC from the Alzheimer’s disease neuroimaging initiative (ADNI) dataset. Experimental results demonstrate that classification accuracy (ACC) of 90.41%, 84.29%, and 65.94% can be achieved for classification of AD vs. NC, pMCI vs. NC, and pMCI vs. sMCI, respectively, indicating the promising performance of the proposed method.
... This is one among the foremost important mechanism to estimate the Alzheimer's where they estimate the tangle and plague of the brain tissues for Alzheimer's analysis [4]. Recently volumetric analysis is used for analyzing manually or semi-automatic techniques using SPM-5 in MATLAB Environment [5,6]. There the neurologist needs to calculate the total volume of the different regions such as white matter, gray matter, CSF and sum together come to conclusion about the stage of the disease. ...
... is indicates that an increase in the network scale affected the number of discriminative features. is result is consistent with the results of similar studies [22,50]. In addition, classification performance analysis showed that a large network significantly improved the accuracy compared with a small network (Table 3). is result is also consistent with the results of previous studies [22][23][24]. ...
Full-text available
Article
In recent years, functional brain network topological features have been widely used as classification features. Previous studies have found that network node scale differences caused by different network parcellation definitions significantly affect the structure of the constructed network and its topological properties. However, we still do not know how network scale differences affect the classification accuracy, performance of classification features, and effectiveness of the feature selection strategy using P values in terms of the machine learning method. This study used five scale parcellations, involving 90, 256, 497, 1003, and 1501 nodes. Three local properties of resting-state functional brain networks were selected (degree, betweenness centrality, and nodal efficiency), and the support vector machine method was used to construct classifiers to identify patients with major depressive disorder. We analyzed the impact of the five scales on classification accuracy. In addition, the effectiveness and redundancy of features obtained by the different scale parcellations were compared. Finally, traditional statistical significance (P value) was verified as a feature selection criterion. The results showed that the feature effectiveness of different scales was similar; in other words, parcellation with more regions did not provide more effective discriminative features. Nevertheless, parcellation with more regions did provide a greater quantity of discriminative features, which led to an improvement in the accuracy of the classification. However, due to the close distance between brain regions, the redundancy of parcellation with more regions was also greater. The traditional P value feature selection strategy is feasible with different scales, but our analysis showed that the traditional P
... Brain atlas selection has been an active field of research in developing MRI-based classification methods where single-atlas or multi-atlas based morphometric representations of brain structures are extracted [4,5,6,7], to name a few. These can serve as prior knowledge to guide predictive models and more recently the training of deep learning architectures for automatic brain labeling [8]. ...
Full-text available
Chapter
Building accurate predictive models to foresee the temporal evolution of diverse medical data representations derived from healthy or disordered brain images will enable a formidable, yet challenging, leap forward in the fields of neuroscience and neuro-disorders. However, such models remain very scarce. Existing landmark works on predicting follow-up medical data from a single observation have a few drawbacks. First, these were developed only for predicting brain shapes or images, while brain network representations remain untapped. Second, the bulk of such models lies in the selection of reliable atlases in the baseline domain, which act as proxies for the follow-up domains where the missing data live. However, current atlas selection strategies for prediction suffer from two major limitations: (i) they are selected based on their proximity to the testing sample using a pre-defined distance, which might not be robust to outliers and constrains the locality of the high-dimensional data to a fixed bandwidth, and (ii) atlases are selected independently of one another, which overlooks how the importance of an individual atlas is influenced by all the other atlases in the set. To address these limitations, we propose LINAs, the first framework for predicting brain network evolution from a single timepoint using learning-guided infinite network atlas selection in two steps. First, we learn how to select the best atlases in an unsupervised manner by learning an adjacency graph which encodes the pairwise similarities between all atlases. The relevance score of an atlas is estimated using all possible infinite paths connecting it to other atlases in the set, quantifying its representativeness and centrality. Second, we propose to individualize the atlas score to the testing sample by a supervised re-weighting strategy. Our comprehensive experiments on healthy and disordered brain networks demonstrate the outperformance of LINAs in comparison with its variants as well as state-of-the-art methods. LINAs presents the first step towards building connectome evolution models that can be leveraged for developing precision medicine.
... Answer: Multi atlas based methods have been used for detection of Alzheimer disease in recent years ( [113][114][115]. Multi atlas based morphometry is used for measuring morphometric representations of the same image in different spaces of multiple atlases. ...
Full-text available
Article
Human age prediction is an interesting and applicable issue in different fields. It can be based on various criteria such as face image, DNA methylation, chest plate radiographs, knee radiographs, dental images and etc. Most of the age prediction researches have mainly been based on images. Since the image processing and Machine Learning (ML) techniques have grown up, the investigations were led to use them in age prediction problem. The implementations would be used in different fields, especially in medical applications. Brain Age Estimation (BAE) has attracted more attention in recent years and it would be so helpful in early diagnosis of some neurodegenerative diseases such as Alzheimer, Parkinson, Huntington, etc. BAE is performed on Magnetic Resonance Imaging (MRI) images to compute the brain ages. Studies based on brain MRI shows that there is a relation between accelerated aging and accelerated brain atrophy. This refers to the effects of neurodegenerative diseases on brain structure while making the whole of it older. This paper reviews and summarizes the main approaches for age prediction based on brain MRI images including preprocessing methods, useful tools used in different research works and the estimation algorithms. We categorize the BAE methods based on two factors, first the way of processing MRI images, which includes pixel-based, surface-based, or voxel-based methods and second, the generation of ML algorithms that includes traditional or Deep Learning (DL) methods. The modern techniques as DL methods help MRI based age prediction to get results that are more accurate. In recent years, more precise and statistical ML approaches have been utilized with the help of related tools for simplifying computations and getting accurate results. Pros and cons of each research and the challenges in each work are expressed and some guidelines and deliberations for future research are suggested.
... Neuroimaging techniques are powerful tools in the assessment of many closely related neurodegenerative disorders [15][16][17][18]. Such tools demonstrate in vivo structural abnormalities and provide a better understanding of the disease mechanisms and progression. ...
Full-text available
Article
Spinocerebellar ataxia type 1 is an autosomal dominant disorder caused by a CAG repeat expansion in ATXN1, characterized by progressive cerebellar and extracerebellar symptoms. MRI-based studies in SCA1 focused in the cerebellum and connections, but there are few data about supratentorial/spinal damage and its clinical relevance. We have thus designed this multimodal MRI study to uncover the structural signature of SCA1. To accomplish that, a group of 33 patients and 33 age-and gender-matched healthy controls underwent MRI on a 3T scanner. All patients underwent a comprehensive neurological and neuropsychological evaluation. We correlated the structural findings with the clinical features of the disease. In addition, we evaluated the disease progression looking at differences in SCA1 subgroups defined by disease duration. Ataxia and pyramidal signs were the main symptoms. Neuropsychological evaluation disclosed cognitive impairment in 53% with predominant frontotemporal dysfunction. Gray matter analysis unfolded cortical thinning of primary and associative motor areas with more restricted impairment of deep structures. Deep gray matter atrophy was associated with motor handicap and poor cognition skills. White matter integrity loss was diffuse in the brainstem but restricted in supratentorial structures. Cerebellar cortical thinning was found in multiple areas and correlated not only with motor disability but also with verbal fluency. Spinal cord atrophy correlated with motor handicap. Comparison of MRI findings in disease duration-defined subgroups identified a peculiar pattern of progressive degeneration.
... To ensure the optimal use of the wealth of information present in fcMRI scans, we used multi-atlas based whole-brain fcMRI in the multivariate pattern analysis, which measures functional connectivity of the same image in different spaces of multiple atlases (Min et al., 2014). The first one includes 176 regions of interest (ROIs) based on the 17-functional network parcellation of the human brain (Buckner et al., 2011;Choi et al., 2012;Yeo et al., 2011). ...
Full-text available
Article
Background: A lack of a sufficiently large sample at single sites causes poor generalizability in automatic diagnosis classification of heterogeneous psychiatric disorders such as schizophrenia based on brain imaging scans. Advanced deep learning methods may be capable of learning subtle hidden patterns from high dimensional imaging data, overcome potential site-related variation, and achieve reproducible cross-site classification. However, deep learning-based cross-site transfer classification, despite less imaging site-specificity and more generalizability of diagnostic models, has not been investigated in schizophrenia. Methods: A large multi-site functional MRI sample (n = 734, including 357 schizophrenic patients from seven imaging resources) was collected, and a deep discriminant autoencoder network, aimed at learning imaging site-shared functional connectivity features, was developed to discriminate schizophrenic individuals from healthy controls. Findings: Accuracies of approximately 85·0% and 81·0% were obtained in multi-site pooling classification and leave-site-out transfer classification, respectively. The learned functional connectivity features revealed dysregulation of the cortical-striatal-cerebellar circuit in schizophrenia, and the most discriminating functional connections were primarily located within and across the default, salience, and control networks. Interpretation: The findings imply that dysfunctional integration of the cortical-striatal-cerebellar circuit across the default, salience, and control networks may play an important role in the "disconnectivity" model underlying the pathophysiology of schizophrenia. The proposed discriminant deep learning method may be capable of learning reliable connectome patterns and help in understanding the pathophysiology and achieving accurate prediction of schizophrenia across multiple independent imaging sites.
... The conventional template techniques can be divided into one-template and multi-template approaches [27]. Recently, several studies [6,21,27,30,46,50] have reported that multi-template approaches often achieve greater precision but have more complex computation than single-template approaches. ...
Full-text available
Article
Visual tracking can be particularly interpreted as a process of searching for targets and optimizing the searching. In this paper, we present a novel tracker framework for tracking shaking targets. We formulate the underlying geometrical relevance between a search scope and a target displacement. A uniform sampling among the search scopes is implemented by sliding windows. To alleviate any possible redundant matching, we propose a double-template structure comprising of initial and previous tracking results. The element-wise similarities between a template and its candidates are calculated by jointly using kernel functions which provide a better outlier rejection property. The STC algorithm is used to improve the tracking results by maximizing a confidence map incorporating temporal and spatial context cues about the tracked targets. For better adaptation to appearance variations, we employ a linear interpolation to update the context prior probability of the STC method. Both qualitative and quantitative evaluations are performed on all sequences that contain shaking motions and are selected from the OTB-50 challenging benchmark. The proposed approach is compared with and outperforms 12 state-of-the-art tracking methods on the selected sequences while running on MATLAB without code optimization. We have also performed further experiments on the whole OTB-50 and VOT 2015 datasets. Although the most of sequences in these two datasets do not contain motion blur that this paper is focusing on, the results of our method are still favorable compared with all of the state-of-the-art approaches.
... State-of-the-art machine learning approaches for this purpose either aim at developing feature selection techniques or focus on designing delicate classifiers. The first type usually use sophisticated techniques for feature selection [37], [38], feature learning [39], or feature extraction [40], [41], [42] and then an straightforward classification technique (like SVM) is utilized. The second type develops task-specific classifiers to enhance the classification accuracies, e.g., [43], [44], [45]. ...
... Previous studies have shown that neuroimaging biomarkers are potential predictors of cognitive impairment (Shi et al., 2010;Cuingnet et al., 2011;Davatzikos et al., 2011;Falahati et al., 2014;Trzepacz et al., 2014;Bron et al., 2015;Jung et al., 2016;Lebedeva et al., 2017). Many researchers have developed and implemented machine learning systems which use neuroimaging biomarkers for more accurate identification of individuals with MCI or dementia (Cui et al., 2012a;Shao et al., 2012;Lebedev et al., 2014;Min et al., 2014;Moradi et al., 2015;Yun et al., 2015;Cai et al., 2017;Guo et al., 2017). Early diagnosis is an essential step in the prevention and early treatment of MCI and dementia. ...
Full-text available
Article
Amnestic MCI (aMCI) and non-amnestic MCI (naMCI) are considered to differ in etiology and outcome. Accurately classifying MCI into meaningful subtypes would enable early intervention with targeted treatment. In this study, we employed structural magnetic resonance imaging (MRI) for MCI subtype classification. This was carried out in a sample of 184 community-dwelling individuals (aged 73–85 years). Cortical surface based measurements were computed from longitudinal and cross-sectional scans. By introducing a feature selection algorithm, we identified a set of discriminative features, and further investigated the temporal patterns of these features. A voting classifier was trained and evaluated via 10 iterations of cross-validation. The best classification accuracies achieved were: 77% (naMCI vs. aMCI), 81% (aMCI vs. cognitively normal (CN)) and 70% (naMCI vs. CN). The best results for differentiating aMCI from naMCI were achieved with baseline features. Hippocampus, amygdala and frontal pole were found to be most discriminative for classifying MCI subtypes. Additionally, we observed the dynamics of classification of several MRI biomarkers. Learning the dynamics of atrophy may aid in the development of better biomarkers, as it may track the progression of cognitive impairment.
Article
Brain graphs are powerful representations to explore the biological roadmaps of the human brain in its healthy and disordered states. Recently, a few graph neural networks (GNNs) have been designed for brain connectivity synthesis and diagnosis. However, such non-Euclidean deep learning architectures might fail to capture the neural interactions between different brain regions as they are trained without guidance from any prior biological template –i.e., template-free learning. Here we assume that using a population-driven brain connectional template (CBT) that captures well the connectivity patterns fingerprinting a given brain state (e.g., healthy) can better guide the GNN training in its downstream learning task such as classification or regression. To this aim we design a plug-in graph registration network (GRN) that can be coupled with any conventional graph neural network (GNN) so as to boost its learning accuracy and generalizability to unseen samples. Our GRN is a graph generative adversarial network (gGAN), which registers brain graphs to a prior CBT. Next, the registered brain graphs are used to train typical GNN models. Our GRN can be integrated into any GNN working in an end-to-end fashion to boost its prediction accuracy. Our experiments showed that GRN remarkably boosted the prediction accuracy of four conventional GNN models across four neurological datasets.
Article
Alzheimer’s disease (AD) is an irreversible neurodegenerative brain disorder that degrades the memory and cognitive ability in elderly people. The main reason for memory loss and reduction in cognitive ability is the structural changes in the brain that occur due to neuronal loss. These structural changes are most conspicuous in the hippocampus, cortex, and grey matter and can be assessed by using neuroimaging techniques viz. Positron Emission Tomography (PET), structural Magnetic Resonance Imaging (MRI) and functional MRI (fMRI), etc. Out of these neuroimaging techniques, structural MRI has evolved as the best technique as it indicates the best soft tissue contrast and high spatial resolution which is important for AD detection. Currently, the focus of researchers is on predicting the conversion of Mild Cognitive Impairment (MCI) into AD. MCI represents the transition state between expected cognitive changes with normal aging and Alzheimer’s disease. Not every MCI patient progresses into Alzheimer’s disease. MCI can develop into stable MCI (sMCI, patients are called non-converters) or into progressive MCI (pMCI, patients are diagnosed as MCI converters). This paper discusses the prognosis of MCI to AD conversion and presents a review of structural MRI-based studies for AD detection. AD detection framework includes feature extraction, feature selection, and classification process. This paper reviews the studies for AD detection based on different feature extraction methods and machine learning algorithms for classification. The performance of various feature extraction methods has been compared and it has been observed that the wavelet transform-based feature extraction method would give promising results for AD classification. The present study indicates that researchers are successful in classifying AD from Normal Controls (NrmC) but, it still requires a lot of work to be done for MCI/ NrmC and MCI/AD, which would help in detecting AD at its early stage.
Chapter
This paper proposes a discriminative transfer feature learning method for MCI conversion prediction using data from the target domain and the auxiliary domain. A transfer component analysis method based on the Maximum Mean Discrepancy (MMD) is proposed at first, which is used to weaken the difference of data distribution between the relevant domain and the target domain. Next, the discriminant optimization term is added to measure the correlation between the sample categories and the sample features of the auxiliary domain, and to improve the inter-class separability of the algorithm. Finally, the support vector machine (SVM) is used to classify MCI patients.KeywordsADMild cognitive impairmentTransfer learningSVM
Article
It has been proven that neuropsychiatric disorders (NDs) can be associated with both structures and functions of brain regions. Thus, data about structures and functions could be usefully combined in a comprehensive analysis. While brain structural MRI (sMRI) images contain anatomic and morphological information about NDs, functional MRI (fMRI) images carry complementary information. However, efficient extraction and fusion of sMRI and fMRI data remains challenging. In this study, we develop an enhanced multi-modal graph convolutional network (MME-GCN) in a binary classification between patients with NDs and healthy controls, based on the fusion of the structural and functional graphs of the brain region. First, based on the same brain atlas, we construct structural and functional graphs from sMRI and fMRI data, respectively. Second, we use machine learning to extract important features from the structural graph network. Third, we use these extracted features to adjust the corresponding edge weights in the functional graph network. Finally, we train a multi-layer GCN and use it in binary classification task. MME-GCN achieved 93.71% classification accuracy on the open data set provided by the Consortium for Neuropsychiatric Phenomics. In addition, we analyzed the important features selected from the structural graph and verified them in the functional graph. Using MME-GCN, we found several specific brain connections important to NDs.
Full-text available
Article
Alzheimer’s Disease (AD) is a neurodegenerative disorder characterized by a progressive loss of memory and a general cognitive decline leading to dementia. AD is characterized by changes in the behavior of the genome and can be traced across multiple brain regions and cell types. It is mainly associated with β-amyloid deposits and tau protein misfolding, leading to neurofibrillary tangles. In recent years, however, research has shown that there is a high complexity of mechanisms involved in AD neurophysiology and functional decline enabling its diverse presentation and allowing more questions to arise. In this study, we present a computational approach to facilitate brain region-specific analysis of genes and biological processes involved in the memory process in AD. Utilizing current genetic knowledge we provide a gene set of 265 memory-associated genes in AD, combinations of which can be found co-expressed in 11 different brain regions along with their functional role. The identified genes participate in a spectrum of biological processes ranging from structural and neuronal communication to epigenetic alterations and immune system responses. These findings provide new insights into the molecular background of AD and can be used to bridge the genotype–phenotype gap and allow for new therapeutic hypotheses.
Full-text available
Article
Background An increase in lifespan in our society is a double-edged sword that entails a growing number of patients with neurocognitive disorders, Alzheimer’s disease being the most prevalent. Advances in medical imaging and computational power enable new methods for the early detection of neurocognitive disorders with the goal of preventing or reducing cognitive decline. Computer-aided image analysis and early detection of changes in cognition is a promising approach for patients with mild cognitive impairment, sometimes a prodromal stage of Alzheimer’s disease dementia. Methods We conducted a systematic review following PRISMA guidelines of studies where machine learning was applied to neuroimaging data in order to predict whether patients with mild cognitive impairment might develop Alzheimer’s disease dementia or remain stable. After removing duplicates, we screened 452 studies and selected 116 for qualitative analysis. Results Most studies used magnetic resonance image (MRI) and positron emission tomography (PET) data but also magnetoencephalography. The datasets were mainly extracted from the Alzheimer’s disease neuroimaging initiative (ADNI) database with some exceptions. Regarding the algorithms used, the most common was support vector machine with a mean accuracy of 75.4%, but convolutional neural networks achieved a higher mean accuracy of 78.5%. Studies combining MRI and PET achieved overall better classification accuracy than studies that only used one neuroimaging technique. In general, the more complex models such as those based on deep learning, combined with multimodal and multidimensional data (neuroimaging, clinical, cognitive, genetic, and behavioral) achieved the best performance. Conclusions Although the performance of the different methods still has room for improvement, the results are promising and this methodology has a great potential as a support tool for clinicians and healthcare professionals.
Article
Multimodal fusion of different types of neural image data provides an irreplaceable opportunity to take advantages of complementary cross-modal information that may only partially be contained in single modality. To jointly analyze multimodal data, deep neural networks can be especially useful because many studies have suggested that deep learning strategy is very efficient to reveal complex and non-linear relations buried in the data. However, most deep models, e.g., convolutional neural network and its numerous extensions, can only operate on regular Euclidean data like voxels in 3D MRI. The interrelated and hidden structures that beyond the grid neighbors, such as brain connectivity, may be overlooked. Moreover, how to effectively incorporate neuroscience knowledge into multimodal data fusion with a single deep framework is understudied. In this work, we developed a graph-based deep neural network to simultaneously model brain structure and function in Mild Cognitive Impairment (MCI): the topology of the graph is initialized using structural network (from diffusion MRI) and iteratively updated by incorporating functional information (from functional MRI) to maximize the capability of differentiating MCI patients from elderly normal controls. This resulted in a new connectome by exploring “deep relations” between brain structure and function in MCI patients and we named it as Deep Brain Connectome. Though deep brain connectome is learned individually, it shows consistent patterns of alteration comparing to structural network at group level. With deep brain connectome, our developed deep model can achieve 92.7% classification accuracy on ADNI dataset.
Article
Accurate classification of Alzheimer's disease (AD) and its prodromal stage mild cognitive impairment (MCI) play key roles in computer-assisted intervention for the diagnosis of AD. However, not all features of AD data will lead to a good classification result, because there are always some unrelated and redundant features. To solve this problem, an adaptive LASSO logistic regression model based on particle swarm optimization(PSO-ALLR)is proposed. This algorithm consists of two stages. In the first stage, the particle swarm optimization (PSO) algorithm is used for global search to remove redundant features and reduces the computational time for the later stage. In the second stage, the adaptive LASSO serves as a local search to select the most relevant features for AD classification.We evaluate the performance of the proposed method on 197 subjects from the baseline MRI data of ADNI database. The proposed method achieves a classification accuracy of 96.27%, 84.81%, and 76.13%, for AD vs. HC, MCI vs. HC, and cMCI vs. sMCI, respectively.
Full-text available
Preprint
Background: Increase in life-span in our society is a double-edged sword that entails a growing number of patients with neurocognitive disorders, Alzheimer’s disease being the most prevalent. Advances in medical imaging and computational power, enable new methods for early detection of neurocognitive disorders with the goal of preventing or reducing cognitive decline. Computer-aided image analysis and early detection of changes in cognition is a promising approach for patients with mild cognitive impairment, sometimes a prodromal stage of Alzheimer’s disease. Methods: We conducted a systematic review following PRISMA guidelines of studies where Machine Learning was applied to neuroimaging data in order to predict the progression from Mild Cognitive Impairment to Alzheimer’s disease. After removing duplicates, we screened 159 studies and selected 47 for a qualitative analysis. Results: Most studies used Magnetic Resonance Image and Positron Emission Tomography data but also Magnetoencephalography. The datasets were mainly extracted from the Alzheimer’s disease Neuroimage Initiative (ADNI) database with some exceptions. Regarding the algorithms used, the most common were support vector machines, but more complex models such as Deep Learning, combined with multimodal and multidimensional data (neuroimaging, clinical, cognitive, biological, and behavioral) achieved the best performance. Conclusions: Although performance of the different models still has room for improvement, the results are promising and this methodology has a great potential as a support tool for clinicians and healthcare professionals.
Article
Accurate prediction of high-risk group who may convert to Alzheimer’s disease (AD) patients is critical for the future treatment of patients. Recently, logistic regression is used for the early diagnosis of AD. However, due to the high-dimensional small sample characteristics of AD data, this brings difficulties to logistic regression-aided diagnosis. To solve the problem, in this paper, we propose sparse logistic regression with the generalized elastic net for the early diagnosis of AD. The generalized elastic net is composed of Lp regularization and L2 regularization. The Lp regularization can produce sparse solutions. L2 regularization ensures that the correlated brain regions are in solution. We evaluate our proposed method on 197 subjects from the baseline MRI data of ADNI database. Our proposed method achieves classification accuracy of 96.10, 84.67, and 75.87 %, for AD vs. HC, MCI vs. HC, and cMCI vs. sMCI, respectively. Experimental results show that, compared with previous methods, our proposed method captures distinct brain regions that are significantly related to AD conversion and provides a significant enhancement in AD classification.
Full-text available
Article
Pain sensitivity is highly variable among individuals, and it is clinically important to predict an individual’s pain sensitivity for individualized diagnosis and management of pain. Literature has shown that pain sensitivity is associated with regional structural features of the brain, but it remains unclear whether pain sensitivity is also related to structural brain connectivity. In the present study, we investigated the relationship between pain thresholds and morphological connectivity (MC) inferred from structural MRI based on data of 221 healthy participants. We found that MC was highly predictive of an individual’s pain thresholds and, importantly, it had a better prediction performance than regional structural features. We also identified a number of most predictive MC features and confirmed the crucial role of the prefrontal cortex in the determination of pain sensitivity. These results suggest the potential of using structural MRI-based MC to predict an individual’s pain sensitivity in clinical settings, and hence this study has important implications for diagnosis and treatment of pain.
Article
Alzheimer’s disease is one of the famous causes of death among elderly. Diagnosis of this disease in the early stage is so difficult by conventional methods. Machine learning methods are one of the best choice for improving the accuracy and performance of diagnosis procedure. The heterogeneous dimensions and structure among the data of this disease have complicated the diagnosis process. Therefore proper features are needed to solve this complexity. In this research, proposed method is introduced in two main steps. In the first step, ensemble of pre-trained auto encoder based feature extraction modules are used to generate image feature from 3D input image and in the second step convolutional neural network is used to diagnosis Alzheimer’s disease. Three different classification cases, namely; Alzheimer’s Disease (AD) versus Normal Condition (NC), AD versus Mild Cognitive Impairment (MCI) and MCI versus NC are studied. Obtained results show that accuracy rate for AD/NC, AD/MCI and MCI/NC are 95%, 90% and 92.5%, respectively. Also, for all cases sensitivity and specially sensitivity rates for proposed method confirm that it could be reliable for diagnosis AD in early stage and has less error to detect normal condition.
Article
In this study, we propose a deep-learning network model called the deep multi-kernel auto-encoder clustering network (DMACN) for clustering functional connectivity data for brain diseases. This model is an end-to-end clustering algorithm that can learn potentially advanced features and cluster disease categories. Unlike other auto-encoders, DMACN has an added self-expression layer and standard back-propagation is used to learn the features that are beneficial for clustering brain functional connectivity data. In the self-expression layer, the kernel matrix is constructed to extract effective features and a new loss function is proposed to constrain the clustering portion, which enables the training of a deep neural learning network that tends to cluster. To test the performance of the proposed algorithm, we applied the end-to-end deep unsupervised clustering algorithm to brain connectivity data. We then conducted experiments based on four public brain functional connectivity data sets and our own functional connectivity data set. The DMACN algorithm yielded good results in various evaluations compared with the existing clustering algorithm for brain functional connectivity data, the deep auto-encoder clustering algorithm, and several other relevant clustering algorithms. The deep-learning-based clustering algorithm has great potential for use in the unsupervised recognition of brain diseases.
Full-text available
Article
Background: In recent years, many convolutional neural networks (CNN) have been proposed for the classification of Alzheimer's disease. Due to memory constraints, many of the proposed CNNs work at a 2D slice-level or 3D patch-level. Objective: Here, we propose a subject-level 3D CNN that can extract the neurodegenerative patterns of the whole brain MRI and converted into a probabilistic Dementia score. Methods: We propose an efficient and lightweight subject-level 3D CNN featuring dilated convolutions. We trained our network on the ADNI data on stable Dementia of the Alzheimer's type (sDAT) from stable normal controls (sNC). To comprehensively evaluate the generalizability of our proposed network, we performed four independent tests which includes testing on images from other ADNI individuals at various stages of the dementia, images acquired from other sites (AIBL), images acquired using different protocols (OASIS), and longitudinal images acquired over a short period of time (MIRIAD). Results: We achieved a 5-fold cross-validated balanced accuracy of 88%in differentiating sDAT from sNC, and an overall specificity of 79.5%and sensitivity 79.7%on the entire set of 7,902 independent test images. Conclusion: Independent testing is essential for estimating the generalization ability of the network to unseen data, but is often lacking in studies using CNN for DAT classification. This makes it difficult to compare the performances achieved using different architectures. Our comprehensive evaluation highlighting the competitive performance of our network and potential promise for generalization.
Full-text available
Article
Accurate classification of Alzheimer’s Disease (AD) and its prodromal stage, i.e., mild cognitive impairment (MCI) are critical for the effective treatment of AD. However, compared with AD classification tasks, predicting the conversion of MCI to AD is relatively difficult. as there are only minor differences among MCI groups. What’s more, in brain imaging analysis, the high dimensionality and relatively small number of subjects brings challenges to computer-aided diagnosis of AD and MCI. Many previous researches focused on the identification of imaging biomarkers for AD diagnosis. In this paper, we introduce sparse logistic regression for the early diagnosis of AD. Sparse logistic regression (SLR) uses L1/2 regularization to impose a sparsity constraint on logistic regression. The L1/2 regularization is considered a representative of Lq regularization, where fewer but informative key brain regions are applied for the classification of AD/MCI. We evaluated the SLR on 197 subjects from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database. Experimental results showed that the SLR improves the classification performance of AD/MCI compared other classical methods.
Chapter
A lack of a sufficiently large sample at single sites causes poor generalizability in automatic diagnostic classification of heterogeneous psychiatric disorders based on brain imaging scans. Advanced deep learning methods may be capable of learning subtle hidden patterns from high-dimensional imaging data, overcome potential site-related variation, and achieve reproducible cross-site classification. So it is of significance to examine deep learning-based cross-site transfer classification of brain imaging, which is less imaging site-specific and more generalizable. In this chapter, a large multi-site functional MRI sample (n = 734, including 357 schizophrenic patients from seven imaging resources) was used, and a deep discriminant autoencoder neural network, aimed at learning imaging site-shared functional connectivity patterns, was developed to discriminate schizophrenic patients from healthy controls. The results revealed that accuracies of approximately 85.0% and 81.0% were obtained in multi-site pooling classification and leave-site-out transfer classification, respectively. The learned functional connectivity patterns revealed dysregulation of the cortical-striatal-cerebellar circuit in schizophrenia, and the most discriminating functional connections were primarily located within and across the default, salience, and control networks. The findings imply that dysfunctional integration of the cortical-striatal-cerebellar circuit across the default, salience, and control networks may play an important role in the “disconnectivity” model underlying the pathophysiology of schizophrenia. The proposed discriminant deep learning method may be capable of learning reliable connectome patterns and help in understanding the pathophysiology and achieving accurate prediction of schizophrenia across multiple independent imaging sites.
Chapter
Resting-state functional connectivity magnetic resonance imaging (rs-fcMRI) studies have shown significant group differences in several regions and networks between depressed patients and healthy controls. This chapter conducted multivariate pattern analysis of whole-brain rs-fcMRI in major depression, which can be used to test the feasibility of identifying major depressive individuals from healthy controls. Twenty-four depressed patients and 29 demographically matched healthy volunteers were included in this study. Permutation tests were used to assess classifier performance. The experimental results demonstrate that 94.3% (P < 0.0001) of subjects were correctly classified via leave-one-out cross-validation, including 100% identification of all patients. The majority of the most discriminating functional connections were located within or across the default mode network, affective network, visual cortical areas, and cerebellum, thereby indicating that the disease-related resting-state network alterations may give rise to a portion of the complex of emotional and cognitive disturbances in major depression. Moreover, the amygdala, anterior cingulate cortex, parahippocampal gyrus, and hippocampus, which exhibit high discriminative power in classification, may play important roles in the pathophysiology of this disorder. The current study may shed new light on the pathological mechanism of major depression and suggests that whole-brain rs-fcMRI may provide potential effective biomarkers for its clinical diagnosis.
Chapter
Autism spectrum disorder (ASD) is a group of neurodevelopmental disorder and its diagnosis is still a challenging issue. To handle it, we propose a novel multi-template ensemble classification framework for ASD diagnosis. Specifically, based on different templates, we construct multiple functional connectivity brain networks for each subject using resting-state functional magnetic resonance imaging (rs-fMRI) data and extract features representations from these networks. Then, our auto-weighted adaptive structural learning model can learn the shared similarity matrix by an adaptive process while selecting informative features. In addition, our method can automatically allot optimal weight for each template without extra weights and parameters. Further, an ensemble classification strategy is adopted to get the final diagnosis results. Our extensive experiments conducted on the Autism Brain Imaging Data Exchange (ABIDE) database demonstrate that our method can improve ASD diagnosis performance. Additionally, our method can detect the ASD-related biomarkers for further medical analysis.
Full-text available
Article
Alzheimer's disease is an incurable neurodegenerative disease primarily affecting the elderly population. Efficient automated techniques are needed for early diagnosis of Alzheimers. Many novel approaches are proposed by researchers for classification of Alzheimer's disease. However, to develop more efficient learning techniques, better understanding of the work done on Alzheimers is needed. Here, we provide a review on 165 papers from 2005-2019 using various feature extraction and machine learning techniques. The machine learning techniques are surveyed under three main categories: support vector machine (SVM), artificial neural network (ANN), and deep learning (DL) and ensemble methods. We present a detailed review on these three approaches for Alzheimers with possible future directions.
Article
Background: Optimized abnormalities of individual brain network may allow earlier detection of mild cognitive impairment (MCI) and accurate prediction of its conversion to Alzheimer's disease (AD). Currently, most studies constructed individual networks based on region-to-region correlation without employing multi-region information. In order to develop the potential discriminative power of network and provide supportive evidence for feasibility of individual metabolic network study, we propose a new approach of extracting indirect relation based network features from18F-FDG PET data. New method: Direct relation based individual network is first constructed using Euclidean distance and exponential function. After that, the lattice-close-degree in fuzzy mathematics is applied to reflect region-to-region indirect relation using the direct relations of regions and their common neighbors. The proposed approach has been evaluated on 199 MCI subjects and 166 normal controls (NC) using SVM classifier. Results: The indirect relation based network features significantly promote classification performances in separating MCI from normal controls (NC) and MCI converters from non-converters. When combining indirect relation features with ADAS-cog scores, further improvements can be obtained. Moreover, the discriminative regions we found are consistent with previous studies, indicating the efficacy of our constructed network in identifying correct biomarkers for diagnosing MCI and predicting its conversion. Comparison with existing method(s): More accurate MCI identification of PET data can be achieved with indirect relation based network features. Conclusions: This work provides a new way to investigate brain network from metabolic perspective for accurate identification of MCI.
Full-text available
Article
Recently, there have been great interests for computer-aided diagnosis of Alzheimer’s disease (AD) and its prodromal stage, mild cognitive impairment (MCI). Unlike the previous methods that considered simple low-level features such as gray matter tissue volumes from MRI, and mean signal intensities from PET, in this paper, we propose a deep learning-based latent feature representation with a stacked auto-encoder (SAE). We believe that there exist latent non-linear complicated patterns inherent in the low-level features such as relations among features. Combining the latent information with the original features helps build a robust model in AD/MCI classification, with high diagnostic accuracy. Furthermore, thanks to the unsupervised characteristic of the pre-training in deep learning, we can benefit from the target-unrelated samples to initialize parameters of SAE, thus finding optimal parameters in fine-tuning with the target-related samples, and further enhancing the classification performances across four binary classification problems: AD vs. healthy normal control (HC), MCI vs. HC, AD vs. MCI, and MCI converter (MCI-C) vs. MCI non-converter (MCI-NC). In our experiments on ADNI dataset, we validated the effectiveness of the proposed method, showing the accuracies of 98.8, 90.7, 83.7, and 83.3 % for AD/HC, MCI/HC, AD/MCI, and MCI-C/MCI-NC classification, respectively. We believe that deep learning can shed new light on the neuroimaging data analysis, and our work presented the applicability of this method to brain disease diagnosis.
Full-text available
Article
Accurate and robust brain extraction is a critical step in most neuroimaging analysis pipelines. In particular, for the large-scale multi-site neuroimaging studies involving a significant number of subjects with diverse age and diagnostic groups, accurate and robust extraction of the brain automatically and consistently is highly desirable. In this paper, we introduce population-specific probability maps to guide the brain extraction of diverse subject groups, including both healthy and diseased adult human populations, both developing and aging human populations, as well as non-human primates. Specifically, the proposed method combines an atlas-based approach, for coarse skull-stripping, with a deformable-surface-based approach that is guided by local intensity information and population-specific prior information learned from a set of real brain images for more localized refinement. Comprehensive quantitative evaluations were performed on the diverse large-scale populations of ADNI dataset with over 800 subjects (55∼90 years of age, multi-site, various diagnosis groups), OASIS dataset with over 400 subjects (18∼96 years of age, wide age range, various diagnosis groups), and NIH pediatrics dataset with 150 subjects (5∼18 years of age, multi-site, wide age range as a complementary age group to the adult dataset). The results demonstrate that our method consistently yields the best overall results across almost the entire human life span, with only a single set of parameters. To demonstrate its capability to work on non-human primates, the proposed method is further evaluated using a rhesus macaque dataset with 20 subjects. Quantitative comparisons with popularly used state-of-the-art methods, including BET, Two-pass BET, BET-B, BSE, HWA, ROBEX and AFNI, demonstrate that the proposed method performs favorably with superior performance on all testing datasets, indicating its robustness and effectiveness.
Full-text available
Article
Deformable image registration is a fundamental task in medical image processing. Among its most important applications, one may cite: i) multi-modality fusion, where information acquired by different imaging devices or protocols is fused to facilitate diagnosis and treatment planning; ii) longitudinal studies, where temporal structural or anatomical changes are investigated; and iii) population modeling and statistical atlases used to study normal anatomical variability. In this paper, we attempt to give an overview of deformable registration methods, putting emphasis on the most recent advances in the domain. Additional emphasis has been given to techniques applied to medical images. In order to study image registration methods in depth, their main components are identified and studied independently. The most recent techniques are presented in a systematic fashion. The contribution of this paper is to provide an extensive account of registration techniques in a systematic manner.
Full-text available
Article
The tutorial starts with an overview of the concepts of VC dimension and structural risk minimization. We then describe linear Support Vector Machines (SVMs) for separable and non-separable data, working through a non-trivial example in detail. We describe a mechanical analogy, and discuss when SVM solutions are unique and when they are global. We describe how support vector training can be practically implemented, and discuss in detail the kernel mapping technique which is used to construct SVM solutions which are nonlinear in the data. We show how Support Vector machines can have very large (even infinite) VC dimension by computing the VC dimension for homogeneous polynomial and Gaussian radial basis function kernels. While very high VC dimension would normally bode ill for generalization performance, and while at present there exists no theory which shows that good generalization performance is guaranteed for SVMs, there are several arguments which support the observed high accuracy of SVMs, which we review. Results of some experiments which were inspired by these arguments are also presented. We give numerous examples and proofs of most of the key theorems. There is new material, and I hope that the reader will find that even old material is cast in a fresh light.
Full-text available
Conference Paper
Skull-stripping refers to the separation of brain tissue from non-brain tissue, such as the scalp, skull, and dura. In large-scale studies involving a significant number of subjects, a fully automatic method is highly desirable, since manual skull-stripping requires tremendous human effort and can be inconsistent even after sufficient training. We propose in this paper a robust and effective method that is capable of skull-stripping a large number of images accurately with minimal dependence on the parameter setting. The key of our method involves an initial skull-stripping by co-registration of an atlas, followed by a refinement phase with a surface deformation scheme that is guided by prior information obtained from a set of real brain images. Evaluation based on a total of 831 images, consisting of normal controls (NC) and patients with mild cognitive impairment (MCI) or Alzheimer's Disease (AD), from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database indicates that our method performs favorably at a consistent overall overlap rate of approximately 98% when compared with expert results. The software package will be made available to the public to facilitate neuroimaging studies.
Full-text available
Article
Recently, several high dimensional classification methods have been proposed to automatically discriminate between patients with Alzheimer's disease (AD) or mild cognitive impairment (MCI) and elderly controls (CN) based on T1-weighted MRI. However, these methods were assessed on different populations, making it difficult to compare their performance. In this paper, we evaluated the performance of ten approaches (five voxel-based methods, three methods based on cortical thickness and two methods based on the hippocampus) using 509 subjects from the ADNI database. Three classification experiments were performed: CN vs AD, CN vs MCIc (MCI who had converted to AD within 18 months, MCI converters - MCIc) and MCIc vs MCInc (MCI who had not converted to AD within 18 months, MCI non-converters - MCInc). Data from 81 CN, 67 MCInc, 39 MCIc and 69 AD were used for training and hyperparameters optimization. The remaining independent samples of 81 CN, 67 MCInc, 37 MCIc and 68 AD were used to obtain an unbiased estimate of the performance of the methods. For AD vs CN, whole-brain methods (voxel-based or cortical thickness-based) achieved high accuracies (up to 81% sensitivity and 95% specificity). For the detection of prodromal AD (CN vs MCIc), the sensitivity was substantially lower. For the prediction of conversion, no classifier obtained significantly better results than chance. We also compared the results obtained using the DARTEL registration to that using SPM5 unified segmentation. DARTEL significantly improved six out of 20 classification experiments and led to lower results in only two cases. Overall, the use of feature selection did not improve the performance but substantially increased the computation times.
Full-text available
Article
Here we develop a multi-template analysis for tensor-based morphometry (TBM) which aims to reduce error from the registration step. In conventional TBM, a single template is nonlinearly registered to all images in the study, and the deformation field statistics are computed from the transformations. Using an MRI dataset from 23 monozygotic and 23 dizygotic twin pairs, we instead registered each individual twin image to 9 additional brain templates using a Riemannian fluid algorithm [3]. Average deformation tensors from multiple registrations were computed within each image, using a log-Euclidean framework [1]. To quantify improvements as the number of registration templates increased from 1 to 9, sequential t-tests assessed the significance of any error reduction, as each new template was added. For each number of templates, we also computed two tensor-derived metrics, and maps of the intraclass correlation of local volume differences, to evaluate any power advantages of multi-atlas TBM. 1
Full-text available
Article
The hippocampal formation (HF) is known from pathological and MRI studies to be severely atrophied in established Alzheimer's disease. However, it is unclear when the earliest changes in the HF occur. We performed a longitudinal study of asymptomatic individuals at risk of autosomal dominant familial Alzheimer's disease in order to assess presymptomatic changes in the HF. Seven at risk members of a familial Alzheimer's disease pedigree associated with the amyloid precursor protein 717 valine to glycine mutation underwent serial MR scanning and neuropsychological assessments over 3 years. These assessments were compared with results from 38 normal controls. During the study three at risk subjects became clinically affected. Volumetric measurement of the HF showed that asymmetrical atrophy developed in these subjects before the appearance of symptoms. Verbal and visual memory measures declined in parallel with hippocampal loss. A loss of up to 8% per annum of the volume of the HF occurred in the 2 years over which symptoms first appeared. These findings may have implications for early diagnosis of Alzheimer's disease.
Full-text available
Article
To determine initial locus and rate of degeneration of temporal lobe structures (total lobe, hippocampus and parahippocampus) in preclinical dementia. Postmortem studies suggest that the earliest changes in Alzheimer's disease are neurofibrillary tangle formation in hippocampus and adjacent cortex. MRI volume analysis of temporal lobe structures over time in subjects prior to developing dementia may allow the identification of when these processes begin, the rate they develop, and which areas are key to symptom development. 30 nondemented (NoD), healthy, elderly individuals enrolled in a prospective study of healthy aging evaluated annually over a mean of 42 months. Twelve subjects with subsequent cognitive decline were assigned to the preclinical dementia group (PreD). All 120 annual MRI studies analyzed by volumetric techniques assessed group differences in temporal lobe volumes and rates of brain loss. NoD as well as PreD subjects had significant, time-dependent decreases in hippocampal and parahippocampal volume. Rates of volume loss between the groups did not significantly differ. PreD cases had significantly smaller hippocampi when asymptomatic. Parahippocampal volume did not differ between PreD and NoD cases. Significant time-dependent temporal lobe atrophy was present only in PreD. Hippocampal and parahippocampal atrophy occurs at a similar rate regardless of diagnostic group. Those who develop dementia may have smaller hippocampi to begin with, but become symptomatic because of accelerated loss of temporal lobe volume. Temporal lobe volume loss may mark the beginning of the disease process within six years prior to dementia onset.
Full-text available
Article
A novel approach to correcting for intensity nonuniformity in magnetic resonance (MR) data is described that achieves high performance without requiring a model of the tissue classes present. The method has the advantage that it can be applied at an early stage in an automated data analysis, before a tissue model is available. Described as nonparametric nonuniform intensity normalization (N3), the method is independent of pulse sequence and insensitive to pathological data that might otherwise violate model assumptions. To eliminate the dependence of the field estimate on anatomy, an iterative approach is employed to estimate both the multiplicative bias field and the distribution of the true tissue intensities. The performance of this method is evaluated using both real and simulated MR data.
Article
Thesupport-vector network is a new learning machine for two-group classification problems. The machine conceptually implements the following idea: input vectors are non-linearly mapped to a very high-dimension feature space. In this feature space a linear decision surface is constructed. Special properties of the decision surface ensures high generalization ability of the learning machine. The idea behind the support-vector network was previously implemented for the restricted case where the training data can be separated without errors. We here extend this result to non-separable training data.High generalization ability of support-vector networks utilizing polynomial input transformations is demonstrated. We also compare the performance of the support-vector network to various classical learning algorithms that all took part in a benchmark study of Optical Character Recognition.
Article
The tutorial starts with an overview of the concepts of VC dimension and structural risk minimization. We then describe linear Support Vector Machines (SVMs) for separable and non-separable data, working through a non-trivial example in detail. We describe a mechanical analogy, and discuss when SVM solutions are unique and when they are global. We describe how support vector training can be practically implemented, and discuss in detail the kernel mapping technique which is used to construct SVM solutions which are nonlinear in the data. We show how Support Vector machines can have very large (even infinite) VC dimension by computing the VC dimension for homogeneous polynomial and Gaussian radial basis function kernels. While very high VC dimension would normally bode ill for generalization performance, and while at present there exists no theory which shows that good generalization performance is guaranteed for SVMs, there are several arguments which support the observed high accuracy of SVMs, which we review. Results of some experiments which were inspired by these arguments are also presented. We give numerous examples and proofs of most of the key theorems. There is new material, and I hope that the reader will find that even old material is cast in a fresh light.
Article
A novel approach to correcting for intensity nonuniformity in magnetic resonance (MR) data is described that achieves high performance without requiring a model of the tissue classes present. The method has the advantage that it can be applied at an early stage in an automated data analysis, before a tissue model is available. Described as nonparametric nonuniform intensity normalization (N3), the method is independent of pulse sequence and insensitive to pathological data that might otherwise violate model assumptions. To eliminate the dependence of the field estimate on anatomy, an iterative approach is employed to estimate both the multiplicative bias field and the distribution of the true tissue intensities. The performance of this method is evaluated using both real and simulated MR data.
Article
LIBSVM is a library for Support Vector Machines (SVMs). We have been actively developing this package since the year 2000. The goal is to help users to easily apply SVM to their applications. LIBSVM has gained wide popularity in machine learning and many other areas. In this article, we present all implementation details of LIBSVM. Issues such as solving SVM optimization problems theoretical convergence multiclass classification probability estimates and parameter selection are discussed in detail.
Article
The support-vector network is a new learning machine for two-group classification problems. The machine conceptually implements the following idea: input vectors are non-linearly mapped to a very high-dimension feature space. In this feature space a linear decision surface is constructed. Special properties of the decision surface ensures high generalization ability of the learning machine. The idea behind the support-vector network was previously implemented for the restricted case where the training data can be separated without errors. We here extend this result to non-separable training data. High generalization ability of support-vector networks utilizing polynomial input transformations is demonstrated. We also compare the performance of the support-vector network to various classical learning algorithms that all took part in a benchmark study of Optical Character Recognition.
Article
This paper is concerned with the development of entropy-based registration criteria for automated 3D multi-modality medical image alignment. In this application where misalignment can be large with respect to the imaged field of view, invariance to overlap statistics is an important consideration. Current entropy measures are reviewed and