Advancing Personalized Prostate Cancer Therapy Through Hormonal Treatment: Promising Findings

Ibrahim Abdelhalim, Ahmed Alksas, Hossam Magdy Balaha, Mohamed A. Badawy, Mohamed Abou El-Ghar, Norah Saleh Alghamdi, Mohammed Ghazal, Sohail Contractor, Eric Van Bogaert, Dibson D. Gondim, S. R. Silva, Fahmi Khalifa, Ayman El-Baz

Abstract: Personalized prediction of hormonal therapy response in Prostate Cancer (PC) is crucial for planning effective treatment. In this paper, we propose a novel framework to combine MRI imaging, pathology, clinical, and demographic markers, aiming to develop a robust prediction system. The process involves sequential steps: preprocessing, prostate/tumor localization, feature extraction, and classification. Using the Multibranch Multimodality MRI Feature Extractor (M3FE), a deep learning technique, we extract salient information from MRI images. The final step employs a weighted sum fusion algorithm to combine MRI features with other markers. Testing on a dataset of 39 patients demonstrates that the framework effectively predicts hormonal therapy effects on PC with 97.5% sensitivity and 100% specificity. This highlights the potential of using radiomics, which involves the analysis of image features, along with other data sources for the precise prediction of hormonal therapy responses in PC.