Example Target Modality Data for Four Clinical Applications
Single 2D chest radiographs (A), examples of knee extremity radiographs drawn from 2D radiograph series (B), 32 slices from 3D head CT scans (HCT) with and without hemorrhage (C), and 19-channel electroencephalography (EEG) signals with and without evidence of seizure onset (D). Note that while these applications are fundamentally different in both dimensionality and modeling requirements (see Table 1), deep machine learning models supporting each can be rapidly trained using cross-modal data programming.