Fig. 1: Analysis Specifications and Flowchart.
From: Multimodal fusion of brain signals for robust prediction of psychosis transition
(Left) Multiple kernel learning (MKL) is a part of the family of kernel methods. These make a prediction about a test sample based on its similarity (kernel function) to samples seen during training. In multimodal multiple kernel learning, schematized here, a multimodal kernel is learned as a combination of kernels obtained from single modalities, allowing patterns to be learned from each modality to inform the rest. (Right) Predictive models were assessed with a unimodal analysis where each data modality was evaluated individually to predict the given label. Next, multimodal analyses are shown using support vector machines (SVM), and finally, a MKL model. RS resting state, DTI diffusion tensor imaging, FA fractional anisotropy, MD mean diffusivity, CHR clinical high risk.
