What if a computer could warn doctors about bowel cancer long before it becomes deadly? A new artificial intelligence tool is doing exactly that — and it could redefine how millions of patients with ulcerative colitis are monitored worldwide.
Colorectal cancer, also known as bowel cancer, affects the colon and rectum and remains one of the world’s deadliest cancers. It is the third most common cancer globally and the second leading cause of cancer-related deaths. For people living with inflammatory bowel diseases such as ulcerative colitis, the danger is even higher. Studies show their risk of developing colorectal cancer is two to three times greater than the general population.
A major challenge for doctors has been predicting which patients with early abnormal cells — known as dysplasia — will actually go on to develop cancer. Dysplasia itself is not cancer, but it can transform into it over time, especially in people with long-standing, untreated inflammation of the bowel. This uncertainty has forced clinicians to rely on frequent colonoscopies or even preventive surgery, often without clear evidence of who truly needs aggressive treatment.
Now, a study published in Clinical Gastroenterology and Hepatology suggests that artificial intelligence may finally solve this problem. Researchers at the University of California, San Diego, developed a fully automated AI system that reads real patient data from electronic health records, including colonoscopy and pathology reports.
The system analyzed records from more than 55,000 patients in the United States Department of Veterans Affairs healthcare system. It extracted clinical details using large language models and identified key predictors of cancer progression, such as lesion size, inflammation severity, and whether abnormal growths could be completely removed. These indicators were then combined with traditional risk factors to build a comprehensive cancer prediction model.
The results were striking. The AI grouped patients into five distinct risk categories that closely matched real-world outcomes over more than ten years of follow-up. Nearly 99 percent of patients placed in the lowest-risk group did not develop colorectal cancer within two years, suggesting that many routine surveillance colonoscopies may be unnecessary for this group.
Kathleen Curtius, PhD, assistant professor of medicine in the Division of Biomedical Informatics at the UC San Diego School of Medicine and one of the study’s authors, said the tool could reduce anxiety, procedures, and healthcare costs for low-risk patients. She explained that current guidelines advise these patients to return for colonoscopy every two years, but the model shows their short-term cancer risk is extremely low, making longer intervals potentially safe.
The AI tool may also change how doctors approach treatment decisions. For patients with low-grade dysplasia, estimating cancer risk has been difficult, leading to frequent procedures and uncertainty about surgery. By using this model, clinicians could personalize surveillance schedules, reserving intensive monitoring for high-risk patients and reducing unnecessary interventions for others.
Importantly, the model also flagged patients with visible lesions that cannot be safely removed due to size or location. These individuals face a much higher risk of developing cancer than many doctors typically estimate. According to Curtius, this is critical because decisions about major preventive surgery — such as partial or total removal of the colon — depend heavily on the risk numbers given to patients.
The researchers emphasize that the tool is not designed to replace doctors. Instead, it is meant to complement clinical judgment and support shared decision-making between patients and healthcare providers. The technology is already available as a web-based tool for clinicians and could soon be integrated directly into hospital electronic record systems.
While the findings are promising, the researchers stress that the model must be validated in more diverse populations beyond the U.S. veterans’ healthcare system. They also plan to incorporate emerging genetic risk factors into the algorithm to improve accuracy even further.
This breakthrough signals a shift toward truly personalized cancer prevention. By identifying who is safe to wait and who urgently needs intervention, artificial intelligence may help doctors save lives while sparing thousands of patients from unnecessary fear, procedures, and surgery.
Stay connected with Medical News Pakistan
CLICK HERE TO JOIN our WhatsApp Community for the latest health updates, life-saving medical news, and expert guidance straight to your phone.