Automatization of the Prostate Cancer Diagnosis Process Based On Biopsy Results Using the EfficientNet

The paper explores the possibility of automatization of the prostate cancer diagnostics process based on histological image analysis. The EfficientNet-b0 neural network model is used as the main tool for such system. The issue of the possibility of retraining a neural network on data from sources not used in the process of its training is considered, as well as the possibility of coarse segmentation and highlighting areas of interest in images in the absence of a sufficient training base with appropriate markup to solve the segmentation problem. The potential is shown to use a basic model trained on PANDA data from European research centers to classify and segment images collected by IMPP FITZ FTM (Novosibirsk).