Multivariate pattern classification of gray matter pathology in multiple sclerosis

A systematic review of the application of machine-learning algorithms in multiple sclerosis

INTRODUCTION

The applications of artificial intelligence, and in particular automatic learning or "machine learning" (ML), constitute both a challenge and a great opportunity in numerous scientific, technical, and clinical disciplines. Specific applications in the study of multiple sclerosis (MS) have been no exception, and constitute an area of increasing interest in recent years.

OBJECTIVE

We present a systematic review of the application of ML algorithms in MS.

MATERIALS AND METHODS

We used the PubMed search engine, which allows free access to the MEDLINE medical database, to identify studies including the keywords "machine learning" and "multiple sclerosis." We excluded review articles, studies written in languages other than English or Spanish, and studies that were mainly technical and did not specifically apply to MS. The final selection included 76 articles, and 38 were rejected.

CONCLUSIONS

After the review process, we established 4 main applications of ML in MS: 1) classifying MS subtypes; 2) distinguishing patients with MS from healthy controls and individuals with other diseases; 3) predicting progression and response to therapeutic interventions; and 4) other applications. Results found to date have shown that ML algorithms may offer great support for health professionals both in clinical settings and in research into MS.

Discovering functional connectivity features characterizing multiple sclerosis phenotypes using explainable artificial intelligence

Multiple sclerosis (MS) is a neurological condition characterized by severe structural brain damage and by functional reorganization of the main brain networks that try to limit the clinical consequences of structural burden. Resting-state (RS) functional connectivity (FC) abnormalities found in this condition were shown to be variable across different MS phases, according to the severity of clinical manifestations. The article describes a system exploiting machine learning on RS FC matrices to discriminate different MS phenotypes and to identify relevant functional connections for MS stage characterization. To this end, the system exploits some mathematical properties of covariance-based RS FC representation, which can be described by a Riemannian manifold. The classification performance of the proposed framework was significantly above the chance level for all MS phenotypes. Moreover, the proposed system was successful in identifying relevant RS FC alterations contributing to an accurate phenotype classification.

Women in the field of multiple sclerosis: How they contributed to paradigm shifts

History is full of women who made enormous contributions to science. While there is little to no imbalance at the early career stage, a decreasing proportion of women is found as seniority increases. In the multiple sclerosis (MS) field, 44% of first authors and only 35% of senior authors were female. So, in this review, we highlight ground-breaking research done by women in the field of MS, focusing mostly on their work as principal investigators. MS is an autoimmune disorder of the central nervous system (CNS), with evident paradigm shifts in the understating of its pathophysiology. It is known that the immune system becomes overactivated and attacks myelin sheath surrounding axons. The resulting demyelination disrupts the communication signals to and from the CNS, which causes unpredictable symptoms, depending on the neurons that are affected. Classically, MS was reported to cause mostly physical and motor disabilities. However, it is now recognized that cognitive impairment affects more than 50% of the MS patients. Another shifting paradigm was the involvement of gray matter in MS pathology, formerly considered to be a white matter disease. Additionally, the identification of different T cell immune subsets and the mechanisms underlying the involvement of B cells and peripheral macrophages provided a better understanding of the immunopathophysiological processes present in MS. Relevantly, the gut-brain axis, recognized as a bi-directional communication system between the CNS and the gut, was found to be crucial in MS. Indeed, gut microbiota influences not only different susceptibilities to MS pathology, but it can also be modulated in order to positively act in MS course. Also, after the identification of the first microRNA in 1993, the role of microRNAs has been investigated in MS, either as potential biomarkers or therapeutic agents. Finally, concerning MS therapeutical approaches, remyelination-based studies have arisen on the spotlight aiming to repair myelin loss/neuronal connectivity. Altogether, here we emphasize the new insights of remarkable women that have voiced the impact of cognitive impairment, white and gray matter pathology, immune response, and that of the CNS-peripheral interplay on MS diagnosis, progression, and/or therapy efficacy, leading to huge breakthroughs in the MS field.

Multimodal-neuroimaging machine-learning analysis of motor disability in multiple sclerosis

Motor disability is a dominant and restricting symptom in multiple sclerosis, yet its neuroimaging correlates are not fully understood. We apply statistical and machine learning techniques on multimodal neuroimaging data to discriminate between multiple sclerosis patients and healthy controls and to predict motor disability scores in the patients. We examine the data of sixty-four multiple sclerosis patients and sixty-five controls, who underwent the MRI examination and the evaluation of motor disability scales. The modalities used comprised regional fractional anisotropy, regional grey matter volumes, and functional connectivity. For analysis, we employ two approaches: high-dimensional support vector machines run on features selected by Fisher Score (aiming for maximal classification accuracy), and low-dimensional logistic regression on the principal components of data (aiming for increased interpretability). We apply analogous regression methods to predict symptom severity. While fractional anisotropy provides the classification accuracy of 96.1% and 89.9% with both approaches respectively, including other modalities did not bring further improvement. Concerning the prediction of motor impairment, the low-dimensional approach performed more reliably. The first grey matter volume component was significantly correlated (R = 0.28-0.46, p < 0.05) with most clinical scales. In summary, we identified the relationship between both white and grey matter changes and motor impairment in multiple sclerosis. Furthermore, we were able to achieve the highest classification accuracy based on quantitative MRI measures of tissue integrity between patients and controls yet reported, while also providing a low-dimensional classification approach with comparable results, paving the way to interpretable machine learning models of brain changes in multiple sclerosis.

Multiple sclerosis diagnosis and phenotype identification by multivariate classification of in vivo frontal cortex metabolite profiles

Multiple sclerosis (MS) is a heterogeneous autoimmune disease for which diagnosis continues to rely on subjective clinical judgment over a battery of tests. Proton magnetic resonance spectroscopy (¹H MRS) enables the noninvasive in vivo detection of multiple small-molecule metabolites and is therefore in principle a promising means of gathering information sufficient for multiple sclerosis diagnosis and subtype classification. Here we show that supervised classification using ¹H-MRS-visible normal-appearing frontal cortex small-molecule metabolites alone can indeed differentiate individuals with progressive MS from control (held-out validation sensitivity 79% and specificity 68%), as well as between relapsing and progressive MS phenotypes (held-out validation sensitivity 84% and specificity 74%). Post hoc assessment demonstrated the disproportionate contributions of glutamate and glutamine to identifying MS status and phenotype, respectively. Our finding establishes ¹H MRS as a viable means of characterizing progressive multiple sclerosis disease status and paves the way for continued refinement of this method as an auxiliary or mainstay of multiple sclerosis diagnostics.

An Enhanced Multiple Sclerosis Disease Diagnosis via an Ensemble Approach

Multiple Sclerosis (MS) is a disease attacking the central nervous system. According to MS Atlas’s most recent statistics, there are more than 2.8 million people worldwide diagnosed with MS. Recently, studies started to explore machine learning techniques to predict MS using various data. The objective of this paper is to develop an ensemble approach for diagnosis of MS using gene expression profiles, while handling the class imbalance problem associated with the data. A hierarchical ensemble approach employing voting and boosting techniques is proposed. This approach adopts a heterogeneous voting approach using two base learners, random forest and support vector machine. Experiments show that our approach outperforms state-of-the-art methods, with the highest recorded accuracy being 92.81% and 93.5% with BoostFS and DEGs for feature selection, respectively. Conclusively, the proposed approach is able to efficiently diagnose MS using the gene expression profiles that are more relevant to the disease. The approach is not merely an ensemble classifier outperforming previous work; it also identifies differentially expressed genes between normal samples and patients with multiple sclerosis using a genome-wide expression microarray. The results obtained show that the proposed approach is an efficient diagnostic tool for MS.

Machine Learning Approaches in Study of Multiple Sclerosis Disease Through Magnetic Resonance Images

Multiple sclerosis (MS) is one of the most common autoimmune diseases which is commonly diagnosed and monitored using magnetic resonance imaging (MRI) with a combination of clinical manifestations. The purpose of this review is to highlight the main applications of Machine Learning (ML) models and their performance in the MS field using MRI. We reviewed the articles of the last decade and grouped them based on the applications of ML in MS using MRI data into four categories: 1) Automated diagnosis of MS, 2) Prediction of MS disease progression, 3) Differentiation of MS stages, 4) Differentiation of MS from similar disorders.

A systematic review of the application of machine-learning algorithms in multiple sclerosis

INTRODUCTION

The applications of artificial intelligence, and in particular automatic learning or "machine learning" (ML), constitute both a challenge and a great opportunity in numerous scientific, technical, and clinical disciplines. Specific applications in the study of multiple sclerosis (MS) have been no exception, and constitute an area of increasing interest in recent years.

OBJECTIVE

We present a systematic review of the application of ML algorithms in MS.

MATERIALS AND METHODS

We used the PubMed search engine, which allows free access to the MEDLINE medical database, to identify studies including the keywords "machine learning" and "multiple sclerosis." We excluded review articles, studies written in languages other than English or Spanish, and studies that were mainly technical and did not specifically apply to MS. The final selection included 76 articles, and 38 were rejected.

CONCLUSIONS

After the review process, we established 4 main applications of ML in MS: 1) classifying MS subtypes; 2) distinguishing patients with MS from healthy controls and individuals with other diseases; 3) predicting progression and response to therapeutic interventions; and 4) other applications. Results found to date have shown that ML algorithms may offer great support for health professionals both in clinical settings and in research into MS.

Deep learning-Based 3D inpainting of brain MR images

The detailed anatomical information of the brain provided by 3D magnetic resonance imaging (MRI) enables various neuroscience research. However, due to the long scan time for 3D MR images, 2D images are mainly obtained in clinical environments. The purpose of this study is to generate 3D images from a sparsely sampled 2D images using an inpainting deep neural network that has a U-net-like structure and DenseNet sub-blocks. To train the network, not only fidelity loss but also perceptual loss based on the VGG network were considered. Various methods were used to assess the overall similarity between the inpainted and original 3D data. In addition, morphological analyzes were performed to investigate whether the inpainted data produced local features similar to the original 3D data. The diagnostic ability using the inpainted data was also evaluated by investigating the pattern of morphological changes in disease groups. Brain anatomy details were efficiently recovered by the proposed neural network. In voxel-based analysis to assess gray matter volume and cortical thickness, differences between the inpainted data and the original 3D data were observed only in small clusters. The proposed method will be useful for utilizing advanced neuroimaging techniques with 2D MRI data.

Evaluation of machine learning algorithms performance for the prediction of early multiple sclerosis from resting-state FMRI connectivity data

Machine Learning application on clinical data in order to support diagnosis and prognostic evaluation arouses growing interest in scientific community. However, choice of right algorithm to use was fundamental to perform reliable and robust classification. Our study aimed to explore if different kinds of Machine Learning technique could be effective to support early diagnosis of Multiple Sclerosis and which of them presented best performance in distinguishing Multiple Sclerosis patients from control subjects. We selected following algorithms: Random Forest, Support Vector Machine, Naïve-Bayes, K-nearest-neighbor and Artificial Neural Network. We applied the Independent Component Analysis to resting-state functional-MRI sequence to identify brain networks. We found 15 networks, from which we extracted the mean signals used into classification. We performed feature selection tasks in all algorithms to obtain the most important variables. We showed that best discriminant network between controls and early Multiple Sclerosis, was the sensori-motor I, according to early manifestation of motor/sensorial deficits in Multiple Sclerosis. Moreover, in classification performance, Random Forest and Support Vector Machine showed same 5-fold cross-validation accuracies (85.7%) using only this network, resulting to be best approaches. We believe that these findings could represent encouraging step toward the translation to clinical diagnosis and prognosis.

SVM recursive feature elimination analyses of structural brain MRI predicts near-term relapses in patients with clinically isolated syndromes suggestive of multiple sclerosis

•

RFE-SVMs predict future outcome of CIS patients with conservative accuracy estimates between 64.9% and 88.1%.

•

Recursive feature selection improves classification performance compared to using all information.

•

Relevant features include regional WM lesion load and GM density, as well as the type of CIS onset.

•

Cross-validation introduces positive bias on accuracy estimate.

Machine learning classification is an attractive approach to automatically differentiate patients from healthy subjects, and to predict future disease outcomes. A clinically isolated syndrome (CIS) is often the first presentation of multiple sclerosis (MS), but it is difficult at onset to predict who will have a second relapse and hence convert to clinically definite MS. In this study, we thus aimed to distinguish CIS converters from non-converters at onset of a CIS, using recursive feature elimination and weight averaging with support vector machines. We also sought to assess the influence of cohort size and cross-validation methods on the accuracy estimate of the classification.

We retrospectively collected 400 patients with CIS from six European MAGNIMS MS centres. Patients underwent brain MRI at onset of a CIS according to local standard-of-care protocols. The diagnosis of clinically definite MS at one-year follow-up was the standard against which the accuracy of the model was tested. For each patient, we derived MRI-based features, such as grey matter probability, white matter lesion load, cortical thickness, and volume of specific cortical and white matter regions. Features with little contribution to the classification model were removed iteratively through an interleaved sample bootstrapping and feature averaging approach. Classification of CIS outcome at one-year follow-up was performed with 2-fold, 5-fold, 10-fold and leave-one-out cross-validation for each centre cohort independently and in all patients together.

The estimated classification accuracy across centres ranged from 64.9% to 88.1% using 2-fold cross-validation and from 73% to 92.9% using leave-one-out cross-validation. The classification accuracy estimate was higher in single-centre, smaller data sets than in combinations of data sets, being the lowest when all patients were merged together.

Regional MRI features such as WM lesions, grey matter probability in the thalamus and the precuneus or cortical thickness in the cuneus and inferior temporal gyrus predicted the occurrence of a second relapse in patients at onset of a CIS using support vector machines. The increased accuracy estimate of the classification achieved with smaller and single-centre samples may indicate a model bias (overfitting) when data points were limited, but also more homogeneous. We provide an overview of classifier performance from a range of cross-validation schemes to give insight into the variability across schemes. The proposed recursive feature elimination approach with weight averaging can be used both in single- and multi-centre data sets in order to bridge the gap between group-level comparisons and making predictions for individual patients.

Uncovering convolutional neural network decisions for diagnosing multiple sclerosis on conventional MRI using layer-wise relevance propagation

Machine learning-based imaging diagnostics has recently reached or even surpassed the level of clinical experts in several clinical domains. However, classification decisions of a trained machine learning system are typically non-transparent, a major hindrance for clinical integration, error tracking or knowledge discovery. In this study, we present a transparent deep learning framework relying on 3D convolutional neural networks (CNNs) and layer-wise relevance propagation (LRP) for diagnosing multiple sclerosis (MS), the most widespread autoimmune neuroinflammatory disease. MS is commonly diagnosed utilizing a combination of clinical presentation and conventional magnetic resonance imaging (MRI), specifically the occurrence and presentation of white matter lesions in T2-weighted images. We hypothesized that using LRP in a naive predictive model would enable us to uncover relevant image features that a trained CNN uses for decision-making. Since imaging markers in MS are well-established this would enable us to validate the respective CNN model. First, we pre-trained a CNN on MRI data from the Alzheimer's Disease Neuroimaging Initiative (n = 921), afterwards specializing the CNN to discriminate between MS patients (n = 76) and healthy controls (n = 71). Using LRP, we then produced a heatmap for each subject in the holdout set depicting the voxel-wise relevance for a particular classification decision. The resulting CNN model resulted in a balanced accuracy of 87.04% and an area under the curve of 96.08% in a receiver operating characteristic curve. The subsequent LRP visualization revealed that the CNN model focuses indeed on individual lesions, but also incorporates additional information such as lesion location, non-lesional white matter or gray matter areas such as the thalamus, which are established conventional and advanced MRI markers in MS. We conclude that LRP and the proposed framework have the capability to make diagnostic decisions of CNN models transparent, which could serve to justify classification decisions for clinical review, verify diagnosis-relevant features and potentially gather new disease knowledge.

Quantifying deep grey matter atrophy using automated segmentation approaches: A systematic review of structural MRI studies

The deep grey matter (DGM) nuclei of the brain play a crucial role in learning, behaviour, cognition, movement and memory. Although automated segmentation strategies can provide insight into the impact of multiple neurological conditions affecting these structures, such as Multiple Sclerosis (MS), Huntington's disease (HD), Alzheimer's disease (AD), Parkinson's disease (PD) and Cerebral Palsy (CP), there are a number of technical challenges limiting an accurate automated segmentation of the DGM. Namely, the insufficient contrast of T1 sequences to completely identify the boundaries of these structures, as well as the presence of iso-intense white matter lesions or extensive tissue loss caused by brain injury. Therefore in this systematic review, 269 eligible studies were analysed and compared to determine the optimal approaches for addressing these technical challenges. The automated approaches used among the reviewed studies fall into three broad categories, atlas-based approaches focusing on the accurate alignment of atlas priors, algorithmic approaches which utilise intensity information to a greater extent, and learning-based approaches that require an annotated training set. Studies that utilise freely available software packages such as FIRST, FreeSurfer and LesionTOADS were also eligible, and their performance compared. Overall, deep learning approaches achieved the best overall performance, however these strategies are currently hampered by the lack of large-scale annotated data. Improving model generalisability to new datasets could be achieved in future studies with data augmentation and transfer learning. Multi-atlas approaches provided the second-best performance overall, and may be utilised to construct a "silver standard" annotated training set for deep learning. To address the technical challenges, providing robustness to injury can be improved by using multiple channels, highly elastic diffeomorphic transformations such as LDDMM, and by following atlas-based approaches with an intensity driven refinement of the segmentation, which has been done with the Expectation Maximisation (EM) and level sets methods. Accounting for potential lesions should be achieved with a separate lesion segmentation approach, as in LesionTOADS. Finally, to address the issue of limited contrast, R2*, T2* and QSM sequences could be used to better highlight the DGM due to its higher iron content. Future studies could look to additionally acquire these sequences by retaining the phase information from standard structural scans, or alternatively acquiring these sequences for only a training set, allowing models to learn the "improved" segmentation from T1-sequences alone.

Characterizing the Structural Pattern Predicting Medication Response in Herpes Zoster Patients Using Multivoxel Pattern Analysis

Herpes zoster (HZ) can cause a blistering skin rash with severe neuropathic pain. Pharmacotherapy is the most common treatment for HZ patients. However, most patients are usually the elderly or those that are immunocompromised, and thus often suffer from side effects or easily get intractable post-herpetic neuralgia (PHN) if medication fails. It is challenging for clinicians to tailor treatment to patients, due to the lack of prognosis information on the neurological pathogenesis that underlies HZ. In the current study, we aimed at characterizing the brain structural pattern of HZ before treatment with medication that could help predict medication responses. High-resolution structural magnetic resonance imaging (MRI) scans of 14 right-handed HZ patients (aged 61.0 ± 7.0, 8 males) with poor response and 15 (aged 62.6 ± 8.3, 5 males) age- (p = 0.58), gender-matched (p = 0.20) patients responding well, were acquired and analyzed. Multivoxel pattern analysis (MVPA) with a searchlight algorithm and support vector machine (SVM), was applied to identify the spatial pattern of the gray matter (GM) volume, with high predicting accuracy. The predictive regions, with an accuracy higher than 79%, were located within the cerebellum, posterior insular cortex (pIC), middle and orbital frontal lobes (mFC and OFC), anterior and middle cingulum (ACC and MCC), precuneus (PCu) and cuneus. Among these regions, mFC, pIC and MCC displayed significant increases of GM volumes in patients with poor response, compared to those with a good response. The combination of sMRI and MVPA might be a useful tool to explore the neuroanatomical imaging biomarkers of HZ-related pain associated with medication responses.

Characterization of relapsing-remitting multiple sclerosis patients using support vector machine classifications of functional and diffusion MRI data

Multiple Sclerosis patients' clinical symptoms do not correlate strongly with structural assessment done with traditional magnetic resonance images. However, its diagnosis and evaluation of the disease's progression are based on a combination of this imaging analysis complemented with clinical examination. Therefore, other biomarkers are necessary to better understand the disease. In this paper, we capitalize on machine learning techniques to classify relapsing-remitting multiple sclerosis patients and healthy volunteers based on machine learning techniques, and to identify relevant brain areas and connectivity measures for characterizing patients. To this end, we acquired magnetic resonance imaging data from relapsing-remitting multiple sclerosis patients and healthy subjects. Fractional anisotropy maps, structural and functional connectivity were extracted from the scans. Each of them were used as separate input features to construct support vector machine classifiers. A fourth input feature was created by combining structural and functional connectivity. Patients were divided in two groups according to their degree of disability and, together with the control group, three group pairs were formed for comparison. Twelve separate classifiers were built from the combination of these four input features and three group pairs. The classifiers were able to distinguish between patients and healthy subjects, reaching accuracy levels as high as 89% ± 2%. In contrast, the performance was noticeably lower when comparing the two groups of patients with different levels of disability, reaching levels below 63% ± 5%. The brain regions that contributed the most to the classification were the right occipital, left frontal orbital, medial frontal cortices and lingual gyrus. The developed classifiers based on MRI data were able to distinguish multiple sclerosis patients and healthy subjects reliably. Moreover, the resulting classification models identified brain regions, and functional and structural connections relevant for better understanding of the disease.

Structural neuroimaging as clinical predictor: A review of machine learning applications

In this paper, we provide an extensive overview of machine learning techniques applied to structural magnetic resonance imaging (MRI) data to obtain clinical classifiers. We specifically address practical problems commonly encountered in the literature, with the aim of helping researchers improve the application of these techniques in future works. Additionally, we survey how these algorithms are applied to a wide range of diseases and disorders (e.g. Alzheimer's disease (AD), Parkinson's disease (PD), autism, multiple sclerosis, traumatic brain injury, etc.) in order to provide a comprehensive view of the state of the art in different fields.

Neuroanatomical morphometric characterization of sex differences in youth using statistical learning

Exploring neuroanatomical sex differences using a multivariate statistical learning approach can yield insights that cannot be derived with univariate analysis. While gross differences in total brain volume are well-established, uncovering the more subtle, regional sex-related differences in neuroanatomy requires a multivariate approach that can accurately model spatial complexity as well as the interactions between neuroanatomical features. Here, we developed a multivariate statistical learning model using a support vector machine (SVM) classifier to predict sex from MRI-derived regional neuroanatomical features from a single-site study of 967 healthy youth from the Philadelphia Neurodevelopmental Cohort (PNC). Then, we validated the multivariate model on an independent dataset of 682 healthy youth from the multi-site Pediatric Imaging, Neurocognition and Genetics (PING) cohort study. The trained model exhibited an 83% cross-validated prediction accuracy, and correctly predicted the sex of 77% of the subjects from the independent multi-site dataset. Results showed that cortical thickness of the middle occipital lobes and the angular gyri are major predictors of sex. Results also demonstrated the inferential benefits of going beyond classical regression approaches to capture the interactions among brain features in order to better characterize sex differences in male and female youths. We also identified specific cortical morphological measures and parcellation techniques, such as cortical thickness as derived from the Destrieux atlas, that are better able to discriminate between males and females in comparison to other brain atlases (Desikan-Killiany, Brodmann and subcortical atlases).

Combined structural and functional patterns discriminating upper limb motor disability in multiple sclerosis using multivariate approaches

A structural or functional pattern of neuroplasticity that could systematically discriminate between people with impaired and preserved motor performance could help us to understand the brain networks contributing to preservation or compensation of behavior in multiple sclerosis (MS). This study aimed to (1) investigate whether a machine learning-based technique could accurately classify MS participants into groups defined by upper extremity function (i.e. motor function preserved (MP) vs. motor function impaired (MI)) based on their regional grey matter measures (GMM, cortical thickness and deep grey matter volume) and inter-regional functional connection (FC), (2) investigate which features (GMM, FC, or GMM + FC) could classify groups more accurately, and (3) identify the multivariate patterns of GMM and FCs that are most discriminative between MP and MI participants, and between each of these groups and the healthy controls (HCs). With 26 MP, 25 MI, and 21 HCs (age and sex matched) underwent T1-weighted and resting-state functional MRI at 3 T, we applied support vector machine (SVM) based classification to learn discriminant functions indicating regions in which GMM or between which FCs were most discriminative between groups. This study demonstrates that there exist structural and FC patterns sufficient for correct classification of upper limb motor ability of people with MS. The classifier with GMM + FC features yielded the highest accuracy of 85.61 % (p < 0.001) to distinguish between the MS groups using leave-one-out cross-validation. It suggests that a machine-learning approach combining structural and functional features is useful for identifying the specific neural substrates that are necessary and sufficient to preserve motor function among people with MS.

Addressing Confounding in Predictive Models with an Application to Neuroimaging

Understanding structural changes in the brain that are caused by a particular disease is a major goal of neuroimaging research. Multivariate pattern analysis (MVPA) comprises a collection of tools that can be used to understand complex disease efxcfects across the brain. We discuss several important issues that must be considered when analyzing data from neuroimaging studies using MVPA. In particular, we focus on the consequences of confounding by non-imaging variables such as age and sex on the results of MVPA. After reviewing current practice to address confounding in neuroimaging studies, we propose an alternative approach based on inverse probability weighting. Although the proposed method is motivated by neuroimaging applications, it is broadly applicable to many problems in machine learning and predictive modeling. We demonstrate the advantages of our approach on simulated and real data examples.

Control-group feature normalization for multivariate pattern analysis of structural MRI data using the support vector machine

Normalization of feature vector values is a common practice in machine learning. Generally, each feature value is standardized to the unit hypercube or by normalizing to zero mean and unit variance. Classification decisions based on support vector machines (SVMs) or by other methods are sensitive to the specific normalization used on the features. In the context of multivariate pattern analysis using neuroimaging data, standardization effectively up- and down-weights features based on their individual variability. Since the standard approach uses the entire data set to guide the normalization, it utilizes the total variability of these features. This total variation is inevitably dependent on the amount of marginal separation between groups. Thus, such a normalization may attenuate the separability of the data in high dimensional space. In this work we propose an alternate approach that uses an estimate of the control-group standard deviation to normalize features before training. We study our proposed approach in the context of group classification using structural MRI data. We show that control-based normalization leads to better reproducibility of estimated multivariate disease patterns and improves the classifier performance in many cases.

Effect of intramuscular interferon beta-1a on gray matter atrophy in relapsing-remitting multiple sclerosis: A retrospective analysis

BACKGROUND

Changes in gray matter (GM) volume may be a useful measure of tissue loss in multiple sclerosis (MS).

OBJECTIVES

To investigate the rate, patterns, and disability correlates of GM volume change in an MS treatment clinical trial.

METHODS

Patients (n=140) with relapsing-remitting MS were randomized to intramuscular (IM) interferon (IFN) beta-1a or placebo. Treatment effects on GM fraction (GMF) and white matter (WM) fraction (WMF) changes, differences in rates of GMF and WMF change in year one and two on treatment, and differences in atrophy rates by disease progression status were assessed retrospectively.

RESULTS

Significantly less GM atrophy (during year two), but not WM atrophy (at any point), was observed with IM IFN beta-1a compared with placebo. Pseudoatrophy effects were more apparent in WM than in GM; in year one, greater WM volume loss was observed with IM IFN beta-1a than with placebo, whereas GM volume loss was similar between groups. Risk of sustained disability progression was significantly associated with GM, but not WM, atrophy.

CONCLUSIONS

These results suggest that GMF change is more meaningful than WMF as a marker of tissue loss and may be useful to augment whole brain atrophy measurements in MS clinical trials.

Detecting the psychosis prodrome across high-risk populations using neuroanatomical biomarkers

To date, the MRI-based individualized prediction of psychosis has only been demonstrated in single-site studies. It remains unclear if MRI biomarkers generalize across different centers and MR scanners and represent accurate surrogates of the risk for developing this devastating illness. Therefore, we assessed whether a MRI-based prediction system identified patients with a later disease transition among 73 clinically defined high-risk persons recruited at two different early recognition centers. Prognostic performance was measured using cross-validation, independent test validation, and Kaplan-Meier survival analysis. Transition outcomes were correctly predicted in 80% of test cases (sensitivity: 76%, specificity: 85%, positive likelihood ratio: 5.1). Thus, given a 54-month transition risk of 45% across both centers, MRI-based predictors provided a 36%-increase of prognostic certainty. After stratifying individuals into low-, intermediate-, and high-risk groups using the predictor's decision score, the high- vs low-risk groups had median psychosis-free survival times of 5 vs 51 months and transition rates of 88% vs 8%. The predictor's decision function involved gray matter volume alterations in prefrontal, perisylvian, and subcortical structures. Our results support the existence of a cross-center neuroanatomical signature of emerging psychosis enabling individualized risk staging across different high-risk populations. Supplementary results revealed that (1) potentially confounding between-site differences were effectively mitigated using statistical correction methods, and (2) the detection of the prodromal signature considerably depended on the available sample sizes. These observations pave the way for future multicenter studies, which may ultimately facilitate the neurobiological refinement of risk criteria and personalized preventive therapies based on individualized risk profiling tools.

Unraveling the relationship between regional gray matter atrophy and pathology in connected white matter tracts in long-standing multiple sclerosis

INTRODUCTION

Gray matter (GM) atrophy is common in multiple sclerosis (MS), but the relationship with white matter (WM) pathology is largely unknown. Some studies found a co-occurrence in specific systems, but a regional analysis across the brain in different clinical phenotypes is necessary to further understand the disease mechanism underlying GM atrophy in MS. Therefore, we investigated the association between regional GM atrophy and pathology in anatomically connected WM tracts.

METHODS

Conventional and diffusion tensor imaging was performed at 3T in 208 patients with long-standing MS and 60 healthy controls. Deep and cortical GM regions were segmented and quantified, and both lesion volumes and average normal appearing WM fractional anisotropy of their associated tracts were derived using an atlas obtained by probabilistic tractography in the controls. Linear regression was then performed to quantify the amount of regional GM atrophy that can be explained by WM pathology in the connected tract.

RESULTS

MS patients showed extensive deep and cortical GM atrophy. Cortical atrophy was particularly present in frontal and temporal regions. Pathology in connected WM tracts statistically explained both regional deep and cortical GM atrophy in relapsing-remitting (RR) patients, but only deep GM atrophy in secondary-progressive (SP) patients.

CONCLUSION

In RRMS patients, both deep and cortical GM atrophy were associated with pathology in connected WM tracts. In SPMS patients, only regional deep GM atrophy could be explained by pathology in connected WM tracts. This suggests that in SPMS patients cortical GM atrophy and WM damage are (at least partly) independent disease processes.

Classification algorithms with multi-modal data fusion could accurately distinguish neuromyelitis optica from multiple sclerosis

Neuromyelitis optica (NMO) exhibits substantial similarities to multiple sclerosis (MS) in clinical manifestations and imaging results and has long been considered a variant of MS. With the advent of a specific biomarker in NMO, known as anti-aquaporin 4, this assumption has changed; however, the differential diagnosis remains challenging and it is still not clear whether a combination of neuroimaging and clinical data could be used to aid clinical decision-making. Computer-aided diagnosis is a rapidly evolving process that holds great promise to facilitate objective differential diagnoses of disorders that show similar presentations. In this study, we aimed to use a powerful method for multi-modal data fusion, known as a multi-kernel learning and performed automatic diagnosis of subjects. We included 30 patients with NMO, 25 patients with MS and 35 healthy volunteers and performed multi-modal imaging with T1-weighted high resolution scans, diffusion tensor imaging (DTI) and resting-state functional MRI (fMRI). In addition, subjects underwent clinical examinations and cognitive assessments. We included 18 a priori predictors from neuroimaging, clinical and cognitive measures in the initial model. We used 10-fold cross-validation to learn the importance of each modality, train and finally test the model performance. The mean accuracy in differentiating between MS and NMO was 88%, where visible white matter lesion load, normal appearing white matter (DTI) and functional connectivity had the most important contributions to the final classification. In a multi-class classification problem we distinguished between all of 3 groups (MS, NMO and healthy controls) with an average accuracy of 84%. In this classification, visible white matter lesion load, functional connectivity, and cognitive scores were the 3 most important modalities. Our work provides preliminary evidence that computational tools can be used to help make an objective differential diagnosis of NMO and MS.

Predicting outcome in clinically isolated syndrome using machine learning

We aim to determine if machine learning techniques, such as support vector machines (SVMs), can predict the occurrence of a second clinical attack, which leads to the diagnosis of clinically-definite Multiple Sclerosis (CDMS) in patients with a clinically isolated syndrome (CIS), on the basis of single patient's lesion features and clinical/demographic characteristics.

Seventy-four patients at onset of CIS were scanned and clinically reviewed after one and three years. CDMS was used as the gold standard against which SVM classification accuracy was tested. Radiological features related to lesional characteristics on conventional MRI were defined a priori and used in combination with clinical/demographic features in an SVM. Forward recursive feature elimination with 100 bootstraps and a leave-one-out cross-validation was used to find the most predictive feature combinations.

30 % and 44 % of patients developed CDMS within one and three years, respectively. The SVMs correctly predicted the presence (or the absence) of CDMS in 71.4 % of patients (sensitivity/specificity: 77 %/66 %) at 1 year, and in 68 % (60 %/76 %) at 3 years on average over all bootstraps. Combinations of features consistently gave a higher accuracy in predicting outcome than any single feature.

Machine-learning-based classifications can be used to provide an “individualised” prediction of conversion to MS from subjects' baseline scans and clinical characteristics, with potential to be incorporated into routine clinical practice.

•

SVMs predict the presence (or absence) of a second clinical attack in Multiple Sclerosis at 1- and 3-year follow-ups.

•

SVM-based classification reaches 71.4 % accuracy, 77 % sensitivity and 66 % specificity for 1-year follow-up.

•

Combinations of features give a higher accuracy than single features.

Preliminary structural MRI based brain classification of chronic pelvic pain: A MAPP network study

Neuroimaging studies have shown that changes in brain morphology often accompany chronic pain conditions. However, brain biomarkers that are sensitive and specific to chronic pelvic pain (CPP) have not yet been adequately identified. Using data from the Trans-MAPP Research Network, we examined the changes in brain morphology associated with CPP. We used a multivariate pattern classification approach to detect these changes and to identify patterns that could be used to distinguish participants with CPP from age-matched healthy controls. In particular, we used a linear support vector machine (SVM) algorithm to differentiate gray matter images from the 2 groups. Regions of positive SVM weight included several regions within the primary somatosensory cortex, pre-supplementary motor area, hippocampus, and amygdala were identified as important drivers of the classification with 73% overall accuracy. Thus, we have identified a preliminary classifier based on brain structure that is able to predict the presence of CPP with a good degree of predictive power. Our regional findings suggest that in individuals with CPP, greater gray matter density may be found in the identified distributed brain regions, which are consistent with some previous investigations in visceral pain syndromes. Future studies are needed to improve upon our identified preliminary classifier with integration of additional variables and to assess whether the observed differences in brain structure are unique to CPP or generalizable to other chronic pain conditions.

Abnormal changes of multidimensional surface features using multivariate pattern classification in amnestic mild cognitive impairment patients

Previous studies have suggested that amnestic mild cognitive impairment (aMCI) is associated with changes in cortical morphological features, such as cortical thickness, sulcal depth, surface area, gray matter volume, metric distortion, and mean curvature. These features have been proven to have specific neuropathological and genetic underpinnings. However, most studies primarily focused on mass-univariate methods, and cortical features were generally explored in isolation. Here, we used a multivariate method to characterize the complex and subtle structural changing pattern of cortical anatomy in 24 aMCI human participants and 26 normal human controls. Six cortical features were extracted for each participant, and the spatial patterns of brain abnormities in aMCI were identified by high classification weights using a support vector machine method. The classification accuracy in discriminating the two groups was 76% in the left hemisphere and 80% in the right hemisphere when all six cortical features were used. Regions showing high weights were subtle, spatially complex, and predominately located in the left medial temporal lobe and the supramarginal and right inferior parietal lobes. In addition, we also found that the six morphological features had different contributions in discriminating the two groups even for the same region. Our results indicated that the neuroanatomical patterns that discriminated individuals with aMCI from controls were truly multidimensional and had different effects on the morphological features. Furthermore, the regions identified by our method could potentially be useful for clinical diagnosis.

Multivariate lesion-symptom mapping using support vector regression

Lesion analysis is a classic approach to study brain functions. Because brain function is a result of coherent activations of a collection of functionally related voxels, lesion-symptom relations are generally contributed by multiple voxels simultaneously. Although voxel-based lesion-symptom mapping (VLSM) has made substantial contributions to the understanding of brain-behavior relationships, a better understanding of the brain-behavior relationship contributed by multiple brain regions needs a multivariate lesion-symptom mapping (MLSM). The purpose of this artilce was to develop an MLSM using a machine learning-based multivariate regression algorithm: support vector regression (SVR). In the proposed SVR-LSM, the symptom relation to the entire lesion map as opposed to each isolated voxel is modeled using a nonlinear function, so the intervoxel correlations are intrinsically considered, resulting in a potentially more sensitive way to examine lesion-symptom relationships. To explore the relative merits of VLSM and SVR-LSM we used both approaches in the analysis of a synthetic dataset. SVR-LSM showed much higher sensitivity and specificity for detecting the synthetic lesion-behavior relations than VLSM. When applied to lesion data and language measures from patients with brain damages, SVR-LSM reproduced the essential pattern of previous findings identified by VLSM and showed higher sensitivity than VLSM for identifying the lesion-behavior relations. Our data also showed the possibility of using lesion data to predict continuous behavior scores.

MRI-based diagnostic biomarkers for early onset pediatric multiple sclerosis

Currently, it is unclear whether pediatric multiple sclerosis (PMS) is a pathoetiologically homogeneous disease phenotype due to clinical and epidemiological differences between early and late onset PMS (EOPMS and LOPMS). Consequently, the question was raised whether diagnostic guidelines need to be complemented by specific EOPMS markers. To search for such markers, we analyzed cerebral MRI images acquired with standard protocols using computer-based classification techniques. Specifically, we applied classification algorithms to gray (GM) and white matter (WM) tissue probability parameters of small brain regions derived from T2-weighted MRI images of EOPMS patients (onset <12 years), LOPMS patients (onset ≥12 years), and healthy controls (HC). This was done for PMS subgroups matched for disease duration and participant age independently. As expected, maximal diagnostic information for distinguishing PMS patients and HC was found in a periventricular WM area containing lesions (87.1% accuracy, p < 2.2 × 10⁻⁵). MRI-based biomarkers specific for EOPMS were identified in prefrontal cortex. Specifically, a coordinate in middle frontal gyrus contained maximal diagnostic information (77.3%, p = 1.8 × 10⁻⁴). Taken together, we were able to identify biomarkers reflecting pathognomonic processes specific for MS patients with very early onset. Especially GM involvement in the separation between PMS subgroups suggests that conventional MRI contains a richer set of diagnostically informative features than previously assumed.

•

EOPMS can be diagnosed accurately with computer-based classification and T2w-MRI.

•

Separation of EOPMS and HC confirmed the pivotal role of WM lesions for diagnosis.

•

Separation of EOPMS and LOPMS showed that GM variations are also informative.

•

Thus, conventional MRI contains a richer set of biomarkers than assumed so far.

Distinguishing prodromal from first-episode psychosis using neuroanatomical single-subject pattern recognition

BACKGROUND

The at-risk mental state for psychosis (ARMS) and the first episode of psychosis have been associated with structural brain abnormalities that could aid in the individualized early recognition of psychosis. However, it is unknown whether the development of these brain alterations predates the clinical deterioration of at-risk individuals, or alternatively, whether it parallels the transition to psychosis at the single-subject level.

METHODS

We evaluated the performance of an magnetic resonance imaging (MRI)-based classification system in classifying disease stages from at-risk individuals with subsequent transition to psychosis (ARMS-T) and patients with first-episode psychosis (FE). Pairwise and multigroup biomarkers were constructed using the structural MRI data of 22 healthy controls (HC), 16 ARMS-T and 23 FE subjects. The performance of these biomarkers was measured in unseen test cases using repeated nested cross-validation.

RESULTS

The classification accuracies in the HC vs FE, HC vs ARMS-T, and ARMS-T vs FE analyses were 86.7%, 80.7%, and 80.0%, respectively. The neuroanatomical decision functions underlying these discriminative results particularly involved the frontotemporal, cingulate, cerebellar, and subcortical brain structures.

CONCLUSIONS

Our findings suggest that structural brain alterations accumulate at the onset of psychosis and occur even before transition to psychosis allowing for the single-subject differentiation of the prodromal and first-episode stages of the disease. Pattern regression techniques facilitate an accurate prediction of these structural brain dynamics at the early stage of psychosis, potentially allowing for the early recognition of individuals at risk of developing psychosis.

Prediction of post-surgical seizure outcome in left mesial temporal lobe epilepsy

Mesial temporal lobe epilepsy is the most common type of focal epilepsy and in its course often becomes refractory to anticonvulsant pharmacotherapy. A resection of the mesial temporal lobe structures is a promising option in these cases. However, approximately 30% of all patients remain with persistent seizures after surgery. In other words, reliable criteria for patients' outcome prediction are absent. To address this limitation, we investigated pre-surgical brain morphology of patients with unilateral left mesial temporal lobe epilepsy who underwent a selective amygdalohippocampectomy. Using support vector classification, we aimed to predict the post-surgical seizure outcome of each patient based on the pre-surgical T1-weighted structural brain images. Due to morphological gender differences and the evidence that men and women differ in onset, prevalence and symptomology in most neurological diseases, we investigated male and female patients separately. Thus, we benefitted from the capability to validate the reliability of our method in two independent samples. Notably, we were able to accurately predict the individual patients' outcome in the male (94% balanced accuracy) as well as in the female (96% balanced accuracy) group. In the male cohort relatively larger white matter volumes in the favorable as compared to the non-favorable outcome group were identified bilaterally in the cingulum bundle, fronto-occipital fasciculus and both caudate nuclei, whereas the left inferior longitudinal fasciculus showed relatively larger white matter volume in the non-favorable group. While relatively larger white matter volumes in the female cohort in the left inferior and right middle longitudinal fasciculus were associated with the favorable outcome, relatively larger white matter volumes in the non-favorable outcome group were identified bilaterally in the superior longitudinal fasciculi I and II. Here, we observed a clear lateralization and distinction of structures involved in the classification in men as compared to women with men exhibiting more alterations in the hemisphere contralateral to the seizure focus. In conclusion, individual post-surgical outcome predictions based on a single T1-weighted magnetic resonance image seem plausible and may thus support the routine pre-surgical workup of epilepsy patients.

The effect of daclizumab on brain atrophy in relapsing-remitting multiple sclerosis

Daclizumab is a monoclonal antibody that reduces inflammation in multiple sclerosis (MS). Through a retrospective analysis, our objective was to determine whether daclizumab treatment reduces the rate of brain structure atrophy in comparison to a mixture of other disease-modifying therapies (mainly different interferon β preparations). We analyzed MRI examinations (1332 scans from 70 MS cases) obtained between 2000 and 2011 in a single center and processed with an automated brain segmentation method. We used mixed-effects multivariable linear regression models to determine whether a median of 4.3 years of daclizumab therapy in 26 patients altered rates of brain-volume change, controlling for variations in MRI protocol. The control group consisted of 44 patients not treated with daclizumab. We found that supratentorial brain volume declined by 5.17 ml per year (95% confidence limits: 3.58-6.77) off daclizumab therapy. On daclizumab, the annual rate of volume loss decreased to 3.72 ml (p=0.01). The rate of ventricular enlargement decreased from 1.26 to 0.42 ml per year (p<0.001). Focused analysis suggests that reduction in gray matter atrophy rate most likely underlies these results. In summary, in this retrospective analysis, daclizumab therapy substantially decreased the rate of brain atrophy in relapsing-remitting MS in comparison to other disease-modifying therapies, predominantly interferon β.

MRI in multiple sclerosis: a review of the current literature

PURPOSE OF REVIEW

This review summarizes the recent data pertaining to the use of magnetic resonance imaging (MRI) in assessing brain and spinal cord involvement in multiple sclerosis (MS).

RECENT FINDINGS

Using MRI as a tool, investigators have made progress recently in understanding the substrate and mechanisms underlying the development and evolution of focal lesions and diffuse damage in MS. The application of refined MRI sequences has markedly improved the characterization of focal lesions, in particular cortical lesions. Promising improvements have been made to clarify the pathological specificity and sensitivity of MRI techniques by performing combined histopathologic-MRI correlation studies. The use of high-field (3 T) and ultra-high-field (UHF; >3 T) MRI has further facilitated the detection of both gray matter and white matter microstructural damage, and elucidated the topographic relationship of overt damage to venous blood vessels. The development of advanced MRI postprocessing tools has led to additional progress in detecting clinically relevant regional gray matter and white matter damage.

SUMMARY

MRI continues to play a pivotal role in the investigation of MS. Ongoing advances in MRI technology should further expand the current understanding of pathologic disease mechanisms and improve diagnostic, prognostic, and monitoring ability in patients with MS.

Classification of different therapeutic responses of major depressive disorder with multivariate pattern analysis method based on structural MR scans

BACKGROUND

Previous studies have found numerous brain changes in patients with major depressive disorder (MDD), but no neurological biomarker has been developed to diagnose depression or to predict responses to antidepressants. In the present study, we used multivariate pattern analysis (MVPA) to classify MDD patients with different therapeutic responses and healthy controls and to explore the diagnostic and prognostic value of structural neuroimaging data of MDD.

METHODOLOGY/PRINCIPAL FINDINGS

Eighteen patients with treatment-resistant depression (TRD), 17 patients with treatment-sensitive depression (TSD) and 17 matched healthy controls were scanned using structural MRI. Voxel-based morphometry, together with a modified MVPA technique which combined searchlight algorithm and principal component analysis (PCA), was used to classify the subjects with TRD, those with TSD and healthy controls. The results revealed that both gray matter (GM) and white matter (WM) of frontal, temporal, parietal and occipital brain regions as well as cerebellum structures had a high classification power in patients with MDD. The accuracy of the GM and WM that correctly discriminated TRD patients from TSD patients was both 82.9%. Meanwhile, the accuracy of the GM that correctly discriminated TRD or TSD patients from healthy controls were 85.7% and 82.4%, respectively; and the WM that correctly discriminated TRD or TSD patients from healthy controls were 85.7% and 91.2%, respectively.

CONCLUSIONS/SIGNIFICANCE

These results suggest that structural MRI with MVPA might be a useful and reliable method to study the neuroanatomical changes to differentiate patients with MDD from healthy controls and patients with TRD from those with TSD. This method might also be useful to study potential brain regions associated with treatment response in patients with MDD.

Classifying minimally disabled multiple sclerosis patients from resting state functional connectivity

Multiple sclerosis (MS), a variable and diffuse disease affecting white and gray matter, is known to cause functional connectivity anomalies in patients. However, related studies published to-date are post hoc; our hypothesis was that such alterations could discriminate between patients and healthy controls in a predictive setting, laying the groundwork for imaging-based prognosis. Using functional magnetic resonance imaging resting state data of 22 minimally disabled MS patients and 14 controls, we developed a predictive model of connectivity alterations in MS: a whole-brain connectivity matrix was built for each subject from the slow oscillations (<0.11 Hz) of region-averaged time series, and a pattern recognition technique was used to learn a discriminant function indicating which particular functional connections are most affected by disease. Classification performance using strict cross-validation yielded a sensitivity of 82% (above chance at p<0.005) and specificity of 86% (p<0.01) to distinguish between MS patients and controls. The most discriminative connectivity changes were found in subcortical and temporal regions, and contralateral connections were more discriminative than ipsilateral connections. The pattern of decreased discriminative connections can be summarized post hoc in an index that correlates positively (ρ=0.61) with white matter lesion load, possibly indicating functional reorganisation to cope with increasing lesion load. These results are consistent with a subtle but widespread impact of lesions in white matter and in gray matter structures serving as high-level integrative hubs. These findings suggest that predictive models of resting state fMRI can reveal specific anomalies due to MS with high sensitivity and specificity, potentially leading to new non-invasive markers.