Jane M. Rondina

PRoNTo: Pattern Recognition for Neuroimaging Toolbox.

In the past years, mass univariate statistical analyses of neuroimaging data have been complemented by the use of multivariate pattern analyses, especially based on machine learning models. While these allow an increased sensitivity for the detection of spatially distributed effects compared to univariate techniques, they lack an established and accessible software framework. The goal of this work was to build a toolbox comprising all the necessary functionalities for multivariate analyses of neuroimaging data, based on machine learning models. The “Pattern Recognition for Neuroimaging Toolbox” (PRoNTo) is open-source, cross-platform, MATLAB-based and SPM compatible, therefore being suitable for both cognitive and clinical neuroscience research. In addition, it is designed to facilitate novel contributions from developers, aiming to improve the interaction between the neuroimaging and machine learning communities. Here, we introduce PRoNTo by presenting examples of possible research questions that can be addressed with the machine learning framework implemented in PRoNTo, and cannot be easily investigated with mass univariate statistical analysis.

Individualized prediction of illness course at the first psychotic episode: a support vector machine MRI study

Background To date, magnetic resonance imaging (MRI) has made little impact on the diagnosis and monitoring of psychoses in individual patients. In this study, we used a support vector machine (SVM) whole-brain classification approach to predict future illness course at the individual level from MRI data obtained at the first psychotic episode. Method One hundred patients at their first psychotic episode and 91 healthy controls had an MRI scan. Patients were re-evaluated 6.2 years (s.d.=2.3) later, and were classified as having a continuous, episodic or intermediate illness course. Twenty-eight subjects with a continuous course were compared with 28 patients with an episodic course and with 28 healthy controls. We trained each SVM classifier independently for the following contrasts: continuous versus episodic, continuous versus healthy controls, and episodic versus healthy controls. Results At baseline, patients with a continuous course were already distinguishable, with significance above chance level, from both patients with an episodic course (p=0.004, sensitivity=71, specificity=68) and healthy individuals (p=0.01, sensitivity=71, specificity=61). Patients with an episodic course could not be distinguished from healthy individuals. When patients with an intermediate outcome were classified according to the discriminating pattern episodic versus continuous, 74% of those who did not develop other episodes were classified as episodic, and 65% of those who did develop further episodes were classified as continuous (p=0.035). Conclusions We provide preliminary evidence of MRI application in the individualized prediction of future illness course, using a simple and automated SVM pipeline. When replicated and validated in larger groups, this could enable targeted clinical decisions based on imaging data.

SCoRS—A Method Based on Stability for Feature Selection and Mapping in Neuroimaging

Feature selection (FS) methods play two important roles in the context of neuroimaging based classification: potentially increase classification accuracy by eliminating irrelevant features from the model and facilitate interpretation by identifying sets of meaningful features that best discriminate the classes. Although the development of FS techniques specifically tuned for neuroimaging data is an active area of research, up to date most of the studies have focused on finding a subset of features that maximizes accuracy. However, maximizing accuracy does not guarantee reliable interpretation as similar accuracies can be obtained from distinct sets of features. In the current paper we propose a new approach for selecting features: SCoRS (survival count on random subsamples) based on a recently proposed Stability Selection theory. SCoRS relies on the idea of choosing relevant features that are stable under data perturbation. Data are perturbed by iteratively sub-sampling both features (subspaces) and examples. We demonstrate the potential of the proposed method in a clinical application to classify depressed patients versus healthy individuals based on functional magnetic resonance imaging data acquired during visualization of happy faces.

Decoding post-stroke motor function from structural brain imaging

Clinical research based on neuroimaging data has benefited from machine learning methods, which have the ability to provide individualized predictions and to account for the interaction among units of information in the brain. Application of machine learning in structural imaging to investigate diseases that involve brain injury presents an additional challenge, especially in conditions like stroke, due to the high variability across patients regarding characteristics of the lesions. Extracting data from anatomical images in a way that translates brain damage information into features to be used as input to learning algorithms is still an open question. One of the most common approaches to capture regional information from brain injury is to obtain the lesion load per region (i.e. the proportion of voxels in anatomical structures that are considered to be damaged). However, no systematic evaluation has yet been performed to compare this approach with using patterns of voxels (i.e. considering each voxel as a single feature). In this paper we compared both approaches applying Gaussian Process Regression to decode motor scores in 50 chronic stroke patients based solely on data derived from structural MRI. For both approaches we compared different ways to delimit anatomical areas: regions of interest from an anatomical atlas, the corticospinal tract, a mask obtained from fMRI analysis with a motor task in healthy controls and regions selected using lesion-symptom mapping. Our analysis showed that extracting features through patterns of voxels that represent lesion probability produced better results than quantifying the lesion load per region. In particular, from the different ways to delimit anatomical areas compared, the best performance was obtained with a combination of a range of cortical and subcortical motor areas as well as the corticospinal tract. These results will inform the appropriate methodology for predicting long term motor outcomes from early post-stroke structural brain imaging.

Brain regions important for recovery after severe post-stroke upper limb paresis

Background The ability to predict outcome after stroke is clinically important for planning treatment and for stratification in restorative clinical trials. In relation to the upper limbs, the main predictor of outcome is initial severity, with patients who present with mild to moderate impairment regaining about 70% of their initial impairment by 3 months post-stroke. However, in those with severe presentations, this proportional recovery applies in only about half, with the other half experiencing poor recovery. The reasons for this failure to recover are not established although the extent of corticospinal tract damage is suggested to be a contributory factor. In this study, we investigated 30 patients with chronic stroke who had presented with severe upper limb impairment and asked whether it was possible to differentiate those with a subsequent good or poor recovery of the upper limb based solely on a T1-weighted structural brain scan. Methods A support vector machine approach using voxel-wise lesion likelihood values was used to show that it was possible to classify patients as good or poor recoverers with variable accuracy depending on which brain regions were used to perform the classification. Results While considering damage within a corticospinal tract mask resulted in 73% classification accuracy, using other (non-corticospinal tract) motor areas provided 87% accuracy, and combining both resulted in 90% accuracy. Conclusion This proof of concept approach highlights the relative importance of different anatomical structures in supporting post-stroke upper limb motor recovery and points towards methodologies that might be used to stratify patients in future restorative clinical trials.

Measuring abnormal brains: building normative rules in neuroimaging using one-class support vector machines

Pattern recognition methods have demonstrated to be suitable analyses tools to handle the high dimensionality of neuroimaging data. However, most studies combining neuroimaging with pattern recognition methods focus on two-class classification problems, usually aiming to discriminate patients under a specific condition (e.g., Alzheimer’s disease) from healthy controls. In this perspective paper we highlight the potential of the one-class support vector machines (OC-SVM) as an unsupervised or exploratory approach that can be used to create normative rules in a multivariate sense. In contrast with the standard SVM that finds an optimal boundary separating two classes (discriminating boundary), the OC-SVM finds the boundary enclosing a specific class (characteristic boundary). If the OC-SVM is trained with patterns of healthy control subjects, the distance to the boundary can be interpreted as an abnormality score. This score might allow quantification of symptom severity or provide insights about subgroups of patients. We provide an intuitive description of basic concepts in one-class classification, the foundations of OC-SVM, current applications, and discuss how this tool can bring new insights to neuroimaging studies.

Framingham Coronary Heart Disease Risk Score Can be Predicted from Structural Brain Images in Elderly Subjects

Recent literature has presented evidence that cardiovascular risk factors (CVRF) play an important role on cognitive performance in elderly individuals, both those who are asymptomatic and those who suffer from symptoms of neurodegenerative disorders. Findings from studies applying neuroimaging methods have increasingly reinforced such notion. Studies addressing the impact of CVRF on brain anatomy changes have gained increasing importance, as recent papers have reported gray matter loss predominantly in regions traditionally affected in Alzheimer’s disease (AD) and vascular dementia in the presence of a high degree of cardiovascular risk. In the present paper, we explore the association between CVRF and brain changes using pattern recognition techniques applied to structural MRI and the Framingham score (a composite measure of cardiovascular risk largely used in epidemiological studies) in a sample of healthy elderly individuals. We aim to answer the following questions: is it possible to decode (i.e., to learn information regarding cardiovascular risk from structural brain images) enabling individual predictions? Among clinical measures comprising the Framingham score, are there particular risk factors that stand as more predictable from patterns of brain changes? Our main findings are threefold: (i) we verified that structural changes in spatially distributed patterns in the brain enable statistically significant prediction of Framingham scores. This result is still significant when controlling for the presence of the APOE 4 allele (an important genetic risk factor for both AD and cardiovascular disease). (ii) When considering each risk factor singly, we found different levels of correlation between real and predicted factors; however, single factors were not significantly predictable from brain images when considering APOE4 allele presence as covariate. (iii) We found important gender differences, and the possible causes of that finding are discussed.

Stability-Based Multivariate Mapping Using SCoRS

Recently we proposed a feature selection method based on stability theory (SCoRS - Survival Count on Random Subspaces) and showed that the proposed approach was able to improve classification accuracy using different datasets. In the present work we propose: (i) an extension of SCoRS using reproducibility instead of model accuracy as the parameter optimization criterion and (ii) a procedure to estimate the rate of false positive selection associated with the set of features obtained. Our results using the proposed framework showed that, as expected, the optimal parameter was more stable across the cross-validation folds, the spatial map displaying the features selected was less noisy and there was no decrease in classification accuracy. In addition, our results suggest that the estimated false positive rate for the features selected by SCoRS is under 0.05 for both optimization approaches, nevertheless lower when optimizing reproducibility in comparison with the standard optimization approach.

A new feature selection method based on stability theory: exploring parameters space to evaluate classification accuracy in neuroimaging data

Recently we proposed a feature selection method based on stability theory. In the present work we present an evaluation of its performance in different contexts through a grid search performed in a subset of its parameters space. The main contributions of this work are: we show that the method can improve the classification accuracy in relation to the wholebrain in different functional datasets; we evaluate the parameters influence in the results, getting some insight in reasonable ranges of values; and we show that combinations of parameters that yield the best accuracies are stable (i.e., they have low rates of false positive selections).

Selecting the most relevant brain regions to discriminate Alzheimer's disease patients from healthy controls using multiple kernel learning: A comparison across functional and structural imaging modalities and atlases

Background Machine learning techniques such as support vector machine (SVM) have been applied recently in order to accurately classify individuals with neuropsychiatric disorders such as Alzheimers disease (AD) based on neuroimaging data. However, the multivariate nature of the SVM approach often precludes the identification of the brain regions that contribute most to classification accuracy. Multiple kernel learning (MKL) is a sparse machine learning method that allows the identification of the most relevant sources for the classification. By parcelating the brain into regions of interest (ROI) it is possible to use each ROI as a source to MKL (ROI-MKL). Methods We applied MKL to multimodal neuroimaging data in order to: 1) compare the diagnostic performance of ROI-MKL and whole-brain SVM in discriminating patients with AD from demographically matched healthy controls and 2) identify the most relevant brain regions to the classification. We used two atlases (AAL and Brodmanns) to parcelate the brain into ROIs and applied ROI-MKL to structural (T1) MRI, 18F-FDG-PET and regional cerebral blood flow SPECT (rCBF-SPECT) data acquired from the same subjects (20 patients with early AD and 18 controls). In ROI-MKL, each ROI received a weight (ROI-weight) that indicated the regions relevance to the classification. For each ROI, we also calculated whether there was a predominance of voxels indicating decreased or increased regional activity (for 18F-FDG-PET and rCBF-SPECT) or volume (for T1-MRI) in AD patients. Results Compared to whole-brain SVM, the ROI-MKL approach resulted in better accuracies (with either atlas) for classification using 18F-FDG-PET (92.5% accuracy for ROI-MKL versus 84% for whole-brain), but not when using rCBF-SPECT or T1-MRI. Although several cortical and subcortical regions contributed to discrimination, high ROI-weights and predominance of hypometabolism and atrophy were identified specially in medial parietal and temporo-limbic cortical regions. Also, the weight of discrimination due to a pattern of increased voxel-weight values in AD individuals was surprisingly high (ranging from approximately 20% to 40% depending on the imaging modality), located mainly in primary sensorimotor and visual cortices and subcortical nuclei. Conclusion The MKL-ROI approach highlights the high discriminative weight of a subset of brain regions of known relevance to AD, the selection of which contributes to increased classification accuracy when applied to 18F-FDG-PET data. Moreover, the MKL-ROI approach demonstrates that brain regions typically spared in mild stages of AD also contribute substantially in the individual discrimination of AD patients from controls.