Stratifying Prostate Cancer: A Comprehensive Framework for Staging and Treatment Planning Classification

With approximately 1.4 million cases recorded in 2020, prostate cancer ranks second in men and fourth overall, making it one of the world’s leading causes of death. For treatment to be effective, prostate cancer must be diagnosed and staged accurately. Nevertheless, there are drawbacks to the present diagnostic methods, including radiation exposure, unequal accuracy, and false positives. These include PSA blood tests, MRIs, CT scans, and PET scans. This study offers a novel method to improve prostate cancer stage identification and classification by utilizing biopsy pictures and Convolutional Neural Networks to address these issues. Better patient outcomes result from the methodology’s provision of enhanced decision-making tools to physicians for treatment planning. The PANDAs dataset consisting of 11000 publicly available clinical largest cancer datasets was used, roughly 8 times the size of the CAMELYON17 challenge. These images are very high resolution (15360 x 13312) representing different cancer grades, providing a diverse set of tissue samples from various stages of malignancy. Pathologists carefully labelled the images to ensure accurate ground truth for model training. Using the EfficientNet7 model, we tested the system with labels provided by qualified radiologists and biopsy-confirmed data and obtained an accuracy of 84% based on the Gleason grading system. This considerably improves the accuracy of the diagnosis and the effectiveness of the treatment compared to the current approaches. Our method enhances the efficacy of early detection techniques and makes this system feasible for practical application. It also contributes to the domains of medical imaging analysis, tumour classification, and radiation planning.