By Sola Ogundipe

Researchers at Mass General Brigham have developed and tested an AI model of reliably assess the size of prostate tumours.

Using MRI scans from over 700 patients, this model successfully identified and outlined 85 percent of the most aggressive prostate lesions shown in imaging.

Larger tumours, as measured by the AI, correlated with a greater chance of treatment failure and spread of cancer, regardless of other common risk factors. For patients receiving radiation therapy, the AI’s tumour size predictions were more accurate than traditional methods in forecasting metastasis.

According to the findings published in the journal Radiology, the tool may assist doctors in evaluating how aggressive a tumor is, allowing for more personalized treatment strategies and improved guidance for radiation therapy.

A researcher in the Department of Radiation Oncology at Brigham and Women’s Hospital, David D. Yang, noted that AI measurements could enhance precision medicine for prostate cancer by helping clinicians recommend the best treatment based on the cancer’s aggressiveness.

MRI technology has significantly improved prostate cancer diagnosis and is now standard in treatment planning. Although doctors can estimate tumour size from MRI images, these estimates can be subjective and vary between individuals.

To achieve a more uniform measurement method, the researchers trained an AI model on MRI images from 732 patients treated at a single facility. They then assessed whether the AI’s size estimates were linked to treatment outcomes over 5 to 10 years post-diagnosis.

The AI model effectively identified and measured about 85 percent of tumours with a high PI-RADS score, indicating a strong likelihood of significant prostate cancer. The size estimates from the AI also served as a potential indicator of prognosis: larger tumors were linked to a higher chance of cancer recurrence or metastasis, regardless of whether patients underwent surgery or radiation therapy.

Senior author Martin King commented that the AI measurements provide meaningful insights into patient outcomes, helping patients understand their cure chances and the likelihood of their cancer returning or spreading.

Beyond assisting clinicians and patients in assessing cancer aggressiveness, the AI model may also aid radiation oncologists by accurately identifying the tumor’s location for targeted treatment. This method is considerably faster than current approaches that typically take two weeks or more to deliver results, allowing patients to start treatment sooner.