A massive killer, colorectal cancer is the second most common cause of cancer deaths in the United States. Unfortunately, figures are still rising, with a third of all cancers diagnosed as colorectal. New diagnostics and treatments would be welcome in this discipline.
In an attempt to build preventative strategies, clinicians have begun screening for pre-cancerous lesions known as adenomas – benign colonic polyps that can turn malignant over time. And even though screening for these polyps can reduce the mortality rate for colon cancer by over 60 percent – detection rates and, conversely, adenoma miss rates vary significantly, ranging from six percent to 41 percent. Meaning a standardized methodology must be introduced into the field that increases detection and reduces miss rates.
Now, a study from researchers led by Beth Israel Deaconess Medical Center reports a reduction of adenoma miss rates by nearly a third when clinicians used artificial intelligence in conjunction with colonoscopy. The team states their study is the first randomized trial examining the role of a deep-learning detection system in colonoscopy in the USA. And is one of the first trials to investigate the role of computer-aided intervention in any field of medicine. The opensource study is published in the journal Clinical Gastroenterology and Hepatology.
Previous studies have shown that increased adenoma detection during colonoscopy is associated with a decreased risk of interval colon cancer. Interval colon cancer is typically diagnosed within 60 months after a negative colonoscopy. However, adenoma detection rates vary significantly among physicians, with studies indicating that adenoma miss rates may also vary between 6 and 41%.
Several solutions are currently in play to counteract the high adenoma miss rate: Firstly, a second observer during colonoscopy has shown varying benefits in increasing adenoma detection. Secondly, computer-aided detection (CADe)– the use of machine learning or deep learning for lesion detection – has recently been applied successfully during colonoscopy.
Researchers have reviewed colonoscopy still-image and video data, then performed randomized clinical trials in China, Italy, and Japan. But as yet, there are no predictive data on the efficacy of artificially intelligent detection systems in a diverse population of patients in the United States.
The current study assesses whether artificial intelligence can improve colonoscopy quality by reducing the adenoma miss rate in a randomized trial.
The trial encompassed 223 patients undergoing colorectal cancer screening at four medical centers from 2019 – 2020. All patients had the standard high-definition, white light colonoscopy and CADe assisted colonoscopy. But half of the participants were randomized to undergo the routine colonoscopy first, followed immediately by the other procedures. And the other half of the volunteers were randomized to receive the screens in reverse order. The same endoscopist performed all colonoscopies.
Data findings show that the adenoma miss rate was just over twenty percent for the group that underwent CADe colonoscopy before the other screens. Significantly lower than the 34% miss rate among those who received standard high-definition white light colonoscopy first.
Interestingly, the rate of false positives was consistent across both groups, and the number of false positives was compatible with false-positive rates in prior studies. However, it’s worth noting that the reported rate of false positives can change depending on the clinical definition used. And to date, no gold standard has been established for CADe aided colonoscopy false-positive rates.
The lab explains how artificial intelligence will become a crucial tool in the fight against cancer. “Our study demonstrates that computer-aided polyp detection has the potential to decrease variability in colonoscopy quality among providers by reducing the miss rate even for experienced physicians,” said senior author Tyler M. Berzin, MD. “These results suggest that artificial intelligence may be an important tool to help reduce the incidence of colorectal cancer in the U.S. through improvements in screening quality.”
In the future, the researchers state because folds in the colonic wall rather than the visual field may have obscured the missed lesions in the CADe aided colonoscopy – further research combining artificial intelligence with mucosal exposure devices is needed.
Image courtesy of rawpixel.com
Don’t miss the latest discoveries from the health innovator community: