Prostate radiotherapy side effects could be anticipated using the very daily scans already taken during treatment, a new study suggests. By applying AI to these setup images, researchers found patterns that may forecast the risk of future rectal bleeding as early as about one week into therapy, opening the door to adapting treatment to minimize harm without sacrificing cancer control.
Radiotherapy is a common and effective option for prostate cancer, but nearby healthy tissues such as the rectum can receive incidental radiation due to their close proximity. Adaptive radiotherapy aims to revise treatment plans regularly in response to changes in a patient’s anatomy, rather than sticking to a single plan for the entire course. Currently, decisions to modify radiotherapy often overlook subtle tissue changes and the texture patterns—radiomic features—that machines can detect.
In this study, researchers from the University of Edinburgh analyzed daily imaging data from 187 men undergoing prostate radiotherapy. Using machine learning, they explored connections between radiomic features extracted from the images and the development of rectal bleeding within two years after treatment.
Remarkably, the study showed that evaluating scans from the first week of treatment could strongly predict later rectal bleeding. When data from the first three weeks were combined, the predictive accuracy improved further, suggesting an early “warning window” for intervention.
Experts say these findings point to a possible role for radiomic feature tracking as a practical, early tool to guide personalized radiotherapy. If validated in larger studies, such an approach could be incorporated into routine planning and monitoring to help clinicians decide not only when to adjust plans but also how to tailor adjustments to each patient’s risk profile.
However, the researchers caution that turning this concept into standard practice will take time. They call for larger trials and additional automation before this method can influence daily clinical decisions.
Related stories
- Early treatment of slow-growing brain tumors may worsen cognitive function
- New NIH grants support research in oral cancer treatment and pain relief
- Do GLP-1 drugs cause cancer? A new review suggests they may not and could offer protection
The study, funded by Prostate Cancer UK, appears in Physics and Imaging in Radiation Oncology (doi: 10.1016/j.phro.2025.100850). The team includes scientists from the University of Cambridge and The Christie NHS Foundation Trust.
Key quotes
Dr Zhuolin Yang, University of Edinburgh: The central finding is that early treatment imaging holds quantitative clues about later toxicity risk, long before symptoms emerge. This reinforces the idea that predictive biomarkers for adaptive radiotherapy might rely on better use of existing data rather than new scans.
Professor Bill Nailon, Edinburgh Cancer Centre: This work demonstrates a proof-of-concept that imaging collected for beam setup could support future adaptive workflows. Real-world adoption will require further trials and automation before clinicians routinely use it to guide decisions.
Dr Hayley Luxton, Prostate Cancer UK: Radiotherapy remains a cornerstone treatment for prostate cancer. Yet organ movement during treatment can cause collateral damage and side effects that affect men’s lives. Our funding—in partnership with Movember and Garfield Weston Foundation—aims to identify who is most at risk so plans can be adjusted early without compromising cancer control. While larger studies are needed, this represents a meaningful step toward kinder, more personalized radiotherapy.
