SNMMI 2026
Can AI forecast radiation dose before prostate cancer treatment?
Clinical Takeaway: Pretherapy 18F-PSMA PET/CT combined with machine learning may enable clinicians to anticipate tumor and organ radiation dose before 177Lu-PSMA therapy, potentially guiding patient selection and dose optimization.
Optimizing dosing for 177Lu-based radiopharmaceutical therapy in metastatic castration-resistant prostate cancer (mCRPC) remains a key challenge, as clinicians typically rely on post-therapy imaging to estimate radiation exposure. This approach is time- and resource-intensive and limits the ability to individualize treatment upfront. A method to predict tumor response and organ dose before therapy could streamline care, inform patient selection, and reduce toxicity risk—making this an area of growing interest for clinicians using targeted radioligand therapies.
A proof-of-concept study presented at SNMMI 2026 demonstrates that machine learning can use pretherapy 18F-PSMA PET/CT data to predict absorbed radiation doses from 177Lu-PSMA therapy. Investigators analyzed 9 patients with mCRPC, evaluating 57 tumors, 36 salivary glands, and 18 kidneys.
The model incorporated PET uptake metrics, radiomic features, and clinical biomarkers within a mixed-effects framework to account for patient-level variability. Predicted absorbed doses were compared against post-therapy dosimetry obtained after one treatment cycle. The approach showed promising agreement between predicted and observed dose distributions across both tumor and normal tissues, suggesting feasibility for pretherapy dose estimation.
“18F-PSMA PET/CT is already routinely performed and widely available... our study sought to determine if information already available from these scans could guide treatment planning before therapy begins,” said lead author Amit Nautiyal, PhD.
Current dosimetry relies on post-treatment imaging, a resource-intensive process that limits real-time clinical decision-making. By shifting dose estimation earlier in the care pathway, this approach could help identify patients likely to benefit from therapy while flagging those at higher risk for organ toxicity.
Although limited by small sample size, the findings support further validation in larger cohorts. “If validated… this approach may improve patient selection and support better decision-making during pre-treatment assessment,” Nautiyal added.
Ongoing research aims to refine and validate the model in multicenter datasets to enable personalized radiopharmaceutical therapy planning.
Source: Nautiyal A, et al. Machine learning-based pretherapy prediction of tumour and organ absorbed dose in 177Lu radiopharmaceutical therapy using 18F PET/CT radiomics and biomarkers. Presented at the Society of Nuclear Medicine and Molecular Imaging (SNMMI) Annual Meeting; May 2026; Los Angeles, CA. Abstract 262138.