FIGURE 1.

Schematic of classification procedure. Within the leave-one-out cross-validation (LOOCV) loop, tract diffusion data from N-1 subjects are normalized and then recursive feature elimination is applied to select specific datapoints used to train the support vector machine (SVM) model. The model is then tested on the tract data from the remaining subject. The LOOCV loops over all N subjects. Outside of the loop, classifications are summed to evaluate SVM accuracy and area under the curve (AUC) values. Additionally, selected datapoints are summed to determine which regions along the tract are selected by the RFE procedure.