Fig. 6
From: Deep learning reveals disease-specific signatures of white matter pathology in tauopathies
Interpretation of WM pathological features used for disease classification. a UMAP visualization of image patches clustered based on their similarity as perceived by the automated disease classifier (based on the 1024-dimensional output of the penultimate layer of the model; “Methods”). Data points are colored based on their ground truth disease label and representative images of data at different areas within clusters are displayed. b The same UMAP visualization from a with data points instead colored by the average feature values (across aggregates in the corresponding image patch) for area, eccentricity, and minor axis length