New AI Approach Aims to Predict Radiation Dose Before Therapy in Advanced Prostate Cancer

Why It Matters

Accurate pre‑treatment dose prediction can streamline patient selection, reduce unnecessary toxicity, and accelerate adoption of PSMA‑targeted radiopharmaceuticals across oncology practices.

Key Takeaways

• AI model predicts ⁷⁷Lu‑PSMA dose using pre‑therapy ¹⁸F‑PSMA PET/CT.
• Study analyzed 9 mCRPC patients, 57 tumors, 36 glands, 18 kidneys.
• Mixed‑effects machine learning combines uptake, radiomics, biomarkers.
• Pre‑therapy predictions could streamline patient selection and reduce toxicity.
• Larger multi‑center trials planned to validate model for clinical use.

Pulse Analysis

Prostate‑specific membrane antigen (PSMA) radioligand therapy has emerged as a cornerstone for treating metastatic castration‑resistant prostate cancer, yet precise dosimetry remains a bottleneck. Traditional post‑therapy imaging to calculate absorbed dose is labor‑intensive and delays treatment decisions. By repurposing the widely available ¹⁸F‑PSMA PET/CT scans performed before therapy, the new AI‑driven workflow promises to shift dosimetry from a retrospective to a prospective tool, aligning with broader trends toward data‑rich, image‑guided oncology.

The proof‑of‑concept study, presented at SNMMI 2026, applied a mixed‑effects machine‑learning model to a cohort of nine patients, extracting quantitative uptake values, advanced radiomic descriptors, and clinical biomarkers. The model’s predictions of tumor and organ dose correlated closely with actual measurements after the first ⁷⁷Lu‑PSMA cycle, suggesting that pre‑therapy imaging contains sufficient information to forecast therapeutic exposure. This capability could enable clinicians to identify patients who are likely to benefit, adjust activity levels to spare critical organs such as kidneys and salivary glands, and potentially reduce the number of treatment cycles required.

Looking ahead, the research team has outlined a five‑year program to expand the dataset across multiple centers, aiming for robust external validation and regulatory acceptance. If successful, the technology could become a standard component of PSMA‑targeted therapy workflows, driving efficiencies for hospitals and pharmaceutical sponsors alike. Early adopters may gain a competitive edge by offering more personalized, lower‑toxicity treatment regimens, while payers could see cost savings from reduced imaging and adverse‑event management.