Fig. 3

a A standard cytokeratin 7 (K7) immunohistochemical stain of a liver core biopsy. b The black circle illustrates the training region, the red circle the annotation region, both drawn by a pathologist to teach the AI model to recognize liver tissue in the image. Space within the black circle but outside the red circle is considered background and not tissue. c The red area is what the K7-AI model considers liver tissue. d The same K7 stain of the same liver biopsy specimen with e training annotations drawn by a pathologist to teach the AI model the difference between portal areas and liver parenchyma. Within the black training regions, circled green areas represent the portal areas and red-circled areas represent parenchyma. f K7-AI model inference mask, meaning the interpretation of the AI model of the specimen. Accordingly, green areas are what the AI model considers as portal areas and red areas are liver parenchyma. g Notice the multiple cholestatic DAB-positive hepatocytes in this K7-stained biopsy specimen. h Red annotation regions within the training regions (black) are the areas taught to the AI model as K7-positive hepatocytes within the liver parenchyma. i The red areas are what the K7-AI model interprets as K7-positive hepatocytes after training