Fusion analysis of functional MRI data for classification of individuals based on patterns of activation

Abstract

Classification of individuals based on patterns of brain activity observed in functional MRI contrasts may be helpful for diagnosis of neurological disorders. Prior work for classification based on these patterns have primarily focused on using a single contrast, which does not take advantage of complementary information that may be available in multiple contrasts. Where multiple contrasts are used, the objective has been only to identify the joint, distinct brain activity patterns that differ between groups of subjects; not to use the information to classify individuals. Here, we use joint Independent Component Analysis (jICA) within a Support Vector Machine (SVM) classification method, and take advantage of the relative contribution of activation patterns generated from multiple fMRI contrasts to improve classification accuracy. Young (age: 19-26) and older (age: 57-73) adults (16 each) were scanned while listening to noise alone and to speech degraded with noise, half of which contained meaningful context that could be used to enhance intelligibility. Functional contrasts based on these conditions (and a silent baseline condition) were used within jICA to generate spatially independent joint activation sources and their corresponding modulation profiles. Modulation profiles were used within a non-linear SVM framework to classify individuals as young or older. Results demonstrate that a combination of activation maps across the multiple contrasts yielded an area under ROC curve of 0.86, superior to classification resulting from individual contrasts. Moreover, class separability, measured by a divergence criterion, was substantially higher when using the combination of activation maps.

Sample-poor estimation of order and common signal subspace with application to fusion of medical imaging data

“…We see a similar N2\P3 complex in the ERP component generated using the sMRI and ERP data. Note that these discriminative components can be leveraged to classify a new subject as either a patient with schizophrenia or a healthy control [28]. By regressing the discriminative components onto the new subject’s feature data, we obtain a set of subject covarations that can be classified using either k -means or a support vector machine.…”

Quantifying the Interaction and Contribution of Multiple Datasets in Fusion: Application to the Detection of Schizophrenia

“…Joint estimation of the sources, and hence, the determination of the individual components, can then be achieved through the performance of a single ICA on the horizontally concatenated X [ k ] defined as $$boldnormal\underset{\_}{X}=[X^{\left[1\right]},X^{\left[2\right]},\dots ,X^{\left[K\right]}]=boldnormalA[S^{\left[1\right]},S^{\left[2\right]},\dots ,S^{\left[k\right]}]=boldnormalA\underset{\_}{S}.$$This method, referred to as jICA [24], is one of the most popular multitask fMRI data fusion methods, see e.g. , [20], [29]. However, jICA is reliant on the assumption that each dataset has the same mixing matrix, thus it may perform poorly when this is not the case [28].…”

“…Despite its widespread use, jICA is a fairly constrained model, since it assumes that all datasets have identical mixing matrices, and thus may perform poorly when the modeling assumptions are not met [28]. This assumption also means that jICA inherently requires each task to contribute similarly to the result, though this may not always be the case in practice [29], [30]. …”

Multimodal Data Fusion Using Source Separation: Application to Medical Imaging

“…This decision was based on the guidance using the information theoretic criterion as well as evaluation of the stability and quality of the results as discussed in Section III.C in [1]. For this problem, the evaluation is based on the statistical significance of the estimated profiles and the interpretability of the estimated components as in [28], [52]. It was observed that the results were quite similar in terms of t-statistics and the estimation of components that are most significant for orders 10 and 20, whereas the performance started to degrade in terms of tstatistics and the estimated areas started to significantly change at orders 5 and 25.…”