Automated Gleason grading of prostate cancers via deep learning in label-free multiphoton microscopic images

Abstract

In the current clinical care, Gleason grading system based on the architectural pattern of cancerous epithelium in histological images is the most powerful prognostic predictor for prostate cancers (PCa). However, the standard procedure of histological examination often includes complicated tissue fixation and staining, which are time-consuming and may delay the diagnosis and surgery. In this study, the unstained prostate tissues were investigated with multiphoton microscopy (MPM) to produce subcellular-resolution images. And then, a deep learning network (AlexNet) was introduced for automated Gleason grading. We achieved an average accuracy of Gleason grading of 78.1%±3.4% for classification. And the area under the curve (AUC) in test set achieves 0.943 which indicates that the proposed model is effective in Gleason grading. At the end, the heat map was performed to visualize the Gleason score of tumour. Our results suggested that MPM, combined with deep learning method, holds the potential to be used as a real-time clinical diagnostic tool for PCa diagnosis.