Skip to main content
Fig. 1 | Acta Neuropathologica Communications

Fig. 1

From: Deep learning reveals disease-specific signatures of white matter pathology in tauopathies

Fig. 1

Schematic of analysis workflow. a Example images of AT8-stained WSI from AD, PSP, and CBD patients that form the basis of our analysis. b Pathologist annotation of cortex(cyan), white matter (magenta) and background (no color) regions were used to train a deep-learning model to segment these regions in WSI. c Characterization of white matter aggregates: A pathologist trained deep learning model was used to segment aggregates in the white matter, and for each aggregate, multiple features characterizing its size and shape were extracted. We then performed unsupervised analyses to test whether white matter aggregates in WSI from the same disease were more similar than those from different diseases. d Disease classification based on cortex and white matter: Separate deep learning models were trained for the cortex and white matter to predict disease status directly from image patches (without need for any human curation of features) and the performance of these two models was compared and contrasted

Back to article page