High Accuracy and Effectiveness with Deep Neural Networks in Pathological Diagnosis of Prostate Cancer

PURPOSE: Due to the multifocality and heterogeneity, the pathological diagnosis of prostate cancer still faces some challenge. To establish an artificial intelligence (AI) system based on prostate histopathological whole mount (WM) sections to make the pathological diagnosis of prostate cancer. EXPERIMENTAL DESIGN: All WM pathological digital images was marked on the basis of the ISUP Gleason grade group by two pathologists. After image segmentation there were totally 2,138,895 patches, of which 1,646,535 patches were valid for training using modified U-Net as fundamental network architecture. Another 22 pieces of WM images from 12 patients were selected for the result testing. Comparisons of Gleason score distributions between training and testing sets were performed by the chi square test. ROC curves for the binary analysis plotted the true positive rate (sensitivity) versus the false positive rate (1 - specificity). The AUC was calculated as proposed by Obuchowski. RESULTS: 826 pieces of WM sections from 148 patients were assigned to the training set randomly. The value of pixel accuracy of three methods (binary analysis, clinically-oriented analysis and analysis for different ISUP Gleason grade) mentioned above were 96.93%, 95.43% and 93.88%, respectively. The value of frequency weighted IoU were 94.32%, 92.13% and 90.21%, respectively. CONCLUSIONS: The AI system not only identified cancer from non-cancer, but also evaluated the malignant degree by distinguishing the ISUP Gleason grade group with high accuracy at pixel level, which is comparable as a professional pathologist. FUNDING STATEMENT: This study was funded by the National Natural Science Foundation of China (81572519, 81772710 and 81802535), Invigorating Health Care through Science (ZDXKB2016014), Key Project Supported by Medical Science and Technology Development Foundation, Nanjing Department of Health (YKK 18064 and YKK18056). DECLARATION OF INTERESTS: The authors declare no competing interests. ETHICS APPROVAL STATEMENT: All the patients involved in the study were provided signed informed consent, which was approved by the ethics committee of Nanjing Drum Tower Hospital.