Fig. 4: Representative benign (A) and malignant (B–F) referral test set cases that were incorrectly called by SpindleMesoNet or a pathologist. | Modern Pathology

Fig. 4: Representative benign (A) and malignant (B–F) referral test set cases that were incorrectly called by SpindleMesoNet or a pathologist.

From: Deep-learning based classification distinguishes sarcomatoid malignant mesotheliomas from benign spindle cell mesothelial proliferations

Fig. 4

The case in (A) was correctly called benign by SpindleMesoNET, but incorrectly called by one pathologist. Subtle zonation and layering support a benign diagnosis. The cases in (B) and (C) were correctly called malignant by SpindleMesoNET, but were incorrectly called by a pathologist. A diagnosis of malignancy is supported by fat invasion in (B) and short storiform architecture in (C). The cases in (D–F) were incorrectly called benign by SpindleMesoNET and either correctly called malignant by pathologists (D, E) or incorrectly called by one pathologist (F). A diagnosis of malignancy is supported by nodularity in (D) and (F) and by necrosis in (E).

Back to article page