Assessment of deep learning assistance for the pathological diagnosis of gastric cancer

• DOI: 10.1038/s41379-022-01073-z
• Publication Date: 2022-04-08T14:06:22Z
• Authors: Wei Ba; Shuhao Wang; Meixia Shang; Ziyan Zhang; Huan Wu; Chunkai Yu; Ranran Xing; Wenjuan Wang; Lang Wang; Cancheng Liu; Huaiyin Shi; Zhigang Song
• Keywords: Cancer Detection

Previous studies on deep learning (DL) applications in pathology have focused pathologist-versus-algorithm comparisons. However, DL will not replace the breadth and contextual knowledge of pathologists; rather, only through their combination may benefits be achieved. A fully crossed multireader multicase study was conducted to evaluate assistance with pathologists' diagnosis gastric cancer. total 110 whole-slide images (WSI) (50 malignant 60 benign) were interpreted by 16 board-certified pathologists or without assistance, a washout period between sessions. DL-assisted achieved higher area under receiver operating characteristic curve (ROC-AUC) (0.911 vs. 0.863, P = 0.003) than unassisted interpreting WSIs. Pathologists demonstrated sensitivity detection cancer (90.63% 82.75%, 0.010). No significant difference observed specificity (78.23% 79.90%, 0.468). The average review time per WSI shortened (22.68 26.37 second, 0.033). Our results that indeed improved accuracy efficiency further boosted acceptance this new technique.