Multi-voxel pattern analysis of fMRI data predicts clinical symptom severity

Abstract

Multi-voxel pattern analysis (MVPA) has been applied successfully to a variety of fMRI research questions in healthy participants. The full potential of applying MVPA to functional data from patient groups has yet to be fully explored. Our goal in this study was to investigate whether MVPA might yield a sensitive predictor of patient symptoms. We also sought to demonstrate that this benefit can be realized from existing datasets, even when they were not designed with MVPA in mind. We analyzed data from an fMRI study of the neural basis for face processing in individuals with an Autism Spectrum Disorder (ASD), who often show fusiform gyrus hypoactivation when presented with unfamiliar faces, compared to controls. We found reliable correlations between MVPA classification performance and standardized measures of symptom severity that exceeded those observed using a univariate measure; a relation that was robust across variations in ROI definition. A searchlight analysis across the ventral temporal lobes identified regions with relationships between classification performance and symptom severity that were not detected using mean activation. These analyses illustrate that MVPA has the potential to act as a sensitive functional biomarker of patient severity.

Figure 1. Examples of stimuli with associated presentation times

Two of the analyzed runs featured face stimuli in a same vs. different task (top row) and two featured a passive-viewing task (bottom row).

Figure 2. Fusiform regions of interest

Regions for each hemisphere (red) are shown on the total area of all regions (white) at x = 39, y = -46, z = -15. The three approaches to defining the regions of interest included: placing spheres at three sets of FFA coordinates from published studies (top), isolating voxel clusters significantly active to faces in the control group (middle), using the clusters of hypoactivation in the ASD group (bottom). Right and left are reversed by convention.

Figure 3. Classification performance within the fusiform regions of interest for control and ASD participants

The bars reflect classification accuracy using the voxel patterns of each region. The dash-marks within each bar show classification performance when the voxel pattern at each timepoint is replaced by the region’s mean activation. Note: it was not appropriate to analyze the controls’ data in the control-group face activation regions, for reasons discussed in the text.

Figure 4. Scatter plots of face vs. house classification performance against ADOS social scores of the ASD participants

Each plot reflects one of the approaches to defining the fusiform regions: coordinate-defined spheres (left), control-group right face activation (middle) and right hypoactive cluster (right). The y-axis begins at the level of chance.

Figure 5. Searchlights with significant correlations between face vs. house classification performance and ADOS social scores

Each red voxel represents the center of one searchlight with a radius of 2 voxels. Significance (p < 0.01) was determined by permuting the clinical scores. Top row: x = 36, y = -40, z = -17. Bottom row: x = 57, y = -31, z = -21. Right and left are reversed by convention.