This research sought to determine if individuals diagnosed with prostate cancer (PCa) would transition to castration-resistant prostate cancer (CRPC) within 12 months following hormone-based therapy. Ninety-six men with PCa who had complete baseline clinical information and underwent multiparametric magnetic resonance imaging (MRI) between September 2018 and September 2022 were retrospectively analyzed. Participants were categorized according to whether they progressed to CRPC after 12 months of hormonal treatment. A Dense Multimodal Fusion Artificial Intelligence (Dense-MFAI) framework was created, incorporating a squeeze-and-excitation module and spatial pyramid pooling into a DenseNet backbone, further combined with the eXtreme Gradient Boosting (XGBoost) learning algorithm. Model performance was assessed using accuracy, sensitivity, specificity, positive and negative predictive values, receiver operating characteristic (ROC) curves, area under the curve (AUC), and confusion matrices. The Dense-MFAI system reached an accuracy rate of 94.2% and an AUC of 0.945 when predicting PCa progression to CRPC within the 12-month treatment window. Experimental validation revealed that merging radiomic signatures with baseline clinical variables enhanced the model’s predictive capability, emphasizing the advantage of multimodal integration. The proposed Dense-MFAI approach effectively forecasts whether PCa will advance to CRPC, providing clinicians with a data-driven tool for optimizing treatment plans and prognostic evaluations.
