Fig. 3: Top features predicting (above) CHR vs. Controls and (below) CHR-Converters vs. CHR-Nonconverters using the MKL model.
From: Multimodal fusion of brain signals for robust prediction of psychosis transition
Top feature weights are shown for A, D subcortical and cortical structural features; B, E functional connectivity features for which green = L and pink = R; and (C, F; bottom two panels) fractional anisotropy features. For all graphs, warm colors denote positive weights (i.e., a higher feature value tilts the classifier toward predicting a CHR and a CHR-Converter label in the first and second model respectively) and connectivity, and cool colors denote negative weights (lower feature values would tilt the decision toward CHR or CHR-Converter). [F = Frontal; C = Central].
