Researchers from Kyoto University used signal patterns derived from a deep neural network to predict visual features from fMRI scans. This allowed their decoder to accurately identify objects.