A Cell Diagnosis Support System with Providing the Reason for Prediction

In this paper we propose a cellular diagnostic support system for cervical cancer from WSI which presents the reason for diagnosis. First, cytoplasmic and nuclear regions are extracted from cell images using U-2-Net. For the extracted regions, 29 image features were calculated, and machine learning is applied to perform cell classification based on the Bethesda system. The results showed an accuracy of 68.5% in the 6-class classification of cervical cancer.