New AI Model Predicts Radiation Doses for Advanced Prostate Cancer Treatment

A new predictive model has been developed to improve radiation dose estimates in advanced prostate cancer treatment. Researchers used pre-therapy scans to create a tool that could make radionuclide therapy more precise. The study focused on patients with metastatic castration-resistant prostate cancer (mCRPC) undergoing ^177Lu-PSMA therapy. The model analyses data from ^18F-PSMA PET/CT scans taken before treatment begins. It combines multiple inputs, including PET uptake measurements, radiomic features, and clinical biomarkers. By using mixed-effects modelling, the system accounts for both fixed factors—like scan and clinical data—and individual patient variations.

Testing involved nine mCRPC patients, with the model evaluating 57 tumours, 36 salivary glands, and 18 kidneys. Early results show strong accuracy in predicting how radiation doses are absorbed in different tissues. This could help doctors tailor doses to maximise tumour exposure while protecting healthy organs. The research is part of a five-year project to refine the model’s precision. Plans include external validation using patient data from multiple medical centres. If confirmed in larger studies, the approach may reduce the need for extensive post-treatment imaging, speeding up therapy decisions. Beyond prostate cancer, the framework could influence other theranostic treatments. Integrating radiomics allows for personalised planning based on a patient’s unique biology. The model may also assist in selecting suitable candidates for ^177Lu-PSMA therapy by forecasting optimal dosing.

The new model offers a clearer way to predict radiation distribution before treatment starts. If validated in broader trials, it could streamline therapy planning and reduce side effects. The findings may also inspire similar advances in other areas of nuclear medicine.