AI Accelerates Individualized Dosimetry for Prostate Cancer Treatment

Radiopharmaceutical therapy is an injected cancer treatment that delivers radiation through the body while selectively targeting tumors. In prostate cancer, current dosing remains largely uniform despite differences in how patients absorb radiation in tumors and healthy organs. This limits clinicians’ ability to balance treatment intensity against toxicity risk. Researchers have now developed an artificial intelligence tool to support individualized dose assessment after therapy.

The University of Massachusetts Amherst developed DiffuDose to generate patient-specific radiation dose maps for prostate cancer treatment with radiopharmaceutical therapy. The model is designed for personalized dosimetry in radiopharmaceutical therapy. It combines two artificial intelligence modules, with one producing a coarse dose estimate and the other refining that estimate into a full-resolution radiation dose map. The system completed this process in under 23 seconds per patient.

The clinical bottleneck addressed by DiffuDose is the time required for gold-standard dose calculation. Although post-treatment scans can show where a radiopharmaceutical concentrates, the image alone does not show the absorbed radiation dose. Conventional computational dosimetry is accurate, but it can require hours for each patient. DiffuDose matched gold-standard accuracy while substantially reducing processing time.

In testing, DiffuDose was compared with six competing methods. It achieved the best overall quantitative performance and maintained high performance across multiple organs. These included both kidneys and the liver, which are important organs at risk for toxicity during radiopharmaceutical therapy. The work was published in IEEE Transactions on Radiation and Plasma Medical Sciences.

The research involved collaborators from UMass Chan Medical School, Massachusetts General Hospital, and the Institute of Nuclear Medicine in Bethesda, Maryland. Future work includes an emerging collaboration with UMass Chan Medical School to build artificial intelligence models using post-radiopharmaceutical therapy scans and patient blood biomarkers. That work is intended to better characterize how patients respond to therapy.

“Right now, everybody gets the same dose. That essentially leaves the therapy's potential untapped, to the extent that it's suboptimal for a given patient. Measuring how much radiation each tissue actually absorbs is the key to personalizing treatment,” said Joyita Dutta, a professor in the Riccio College of Engineering at UMass Amherst.

Related Links
UMass Amherst