Skip to main content
Fig. 4 | Cancer Imaging

Fig. 4

From: Utility of the deep learning technique for the diagnosis of orbital invasion on CT in patients with a nasal or sinonasal tumor

Fig. 4

Representative cases diagnosed as false-negative and false-positive with the trained CNN model. a A patient with left ethmoid sinus cancer. As the tumor involved the extraocular muscles and obliterated orbital fat with medial orbital wall destruction (white arrow), the final diagnosis by experienced radiologists was determined as invasion-positive. However, the CNN model-based diagnosis was negative (a false negative). b A patient with left maxillary sinus cancer. As a thin membrane structure bordering the tumor and the intraorbital fat could be identified (white arrowhead), the final diagnosis by the experienced radiologists was invasion-negative. However, the CNN model-based diagnosis was positive (false positive)

Back to article page