Weakly Supervised Convolutional Neural Network for Automatic Gleason Grading of Prostate Cancer
Maryam Kamareh, Mohammad Sadegh Helfroush, Kamran Kazemi · 2022
Digital histopathology is based on the analysis of digitized biopsy slides. Prostate cancer is an usual disease among men. Process of analyzing histopathology images and manually determining the Gleason grades by expert pathologists takes a lot of time. Therefore, development of a system based on machine learning can provide an accurate method for grading prostate cancer. But development of these systems are difficult because they need significant amounts of pixel level annotated data. Since the pathologists’ clinical reports often contain only slide-level labels, this type of data is rarely available. Therefore, the development of methods that can learn using only slide-level labels and do not require manual pixel level annotation would be a considerable advance in this field. In this paper, we design a weakly supervised convolutional neural network for Gleason grading in tissue microarrays, without using pixel level annotations. We explored different pre-trained models as a backbone of our models, namely ResNet-50, VGG-19 and MobileNet. We used class-wise data augmentation method to face the imbalance problem in our dataset. The best network architecture was ResNet-50 as backbone. In the test cohort, ResNet-50 as backbone with class-wise data augmentation achieved an accuracy of 80% for the Gleason grades.