Fig. 2: Diagram summarizing the methodology used to create the virtual immunohistochemistry (vIHC) neural network training set. | Modern Pathology

Fig. 2: Diagram summarizing the methodology used to create the virtual immunohistochemistry (vIHC) neural network training set.

From: A machine learning algorithm for simulating immunohistochemistry: development of SOX10 virtual IHC and evaluation on primarily melanocytic neoplasms

Fig. 2

Hematoxylin and eosin (H&E) slides were first prepared from formalin-fixed paraffin-embedded tissue. The H&E slides were first scanned (a) to create digital H&E whole slide images (WSI). The H&E slides were then (b) destained and (c) stained with SOX10 immunohistochemistry (IHC). The resulting IHC slides were scanned (d) to create a digital IHC WSI. The H&E and corresponding IHC WSIs were registered (e), and processed (f) to create a ground-truth mask. In the ground-truth masks, SOX10-positive nuclei are colored yellow and SOX10-negative nuclei are colored green.

Back to article page