Skip to main content
. 2019 Aug;4(8):726–733. doi: 10.1016/j.bpsc.2019.04.005

Figure 1.

Figure 1

Multiple kernel learning (MKL) framework. Training phase: (A) the MKL regression model is trained by providing examples that pair a contrast image from the general linear model (brain patterns) and a clinical score. (B) The MKL framework uses a predefined anatomical template to segment the contrast images into 116 anatomical brain regions. (C) The MKL simultaneously learns the contribution of each region for the decision function (region weights or contribution) and within each region the contribution of each voxel (voxel weights). Testing phase: (D) During the testing phase, a new contrast image (brain patterns) of a test subject is given as input for the MKL model. (E) This contrast image is parcellated using anatomical atlas. (F) The MKL regression model is applied to the segmented contrast image to predict the clinical score. (G) The model performance is evaluated using two metrics to measure the agreement between the predicted and the actual clinical scores: Pearson’s correlation coefficient (r) and mean squared error (MSE). AAL, Automated Anatomical Labeling.