According to the American Cancer Society, colorectal cancer is the third most common cancer diagnosed in both men and women in the United States. Colorectal cancer (CRC) is the second most common cause of cancer deaths when men and women are combined, and in 2021, it’s expected to cause around 52,980 deaths.
Colorectal cancer begins with polyps or other precancerous growths in the large intestine. Like breast cancer, colon screening has become a commonplace practice. Typically, gastroenterologists will administer the screening via a colonoscopy to identify any irregularities in the lining of the colon and rectum.
Although the screening infrastructure around colon health has dramatically improved the rates of early diagnosis and decreased CRC mortality rates, colonoscopies aren’t foolproof. A population-based study from the Huntsman Cancer Institute at the University of Utah found that around 6% of colorectal cancers are diagnosed within three to five years after the patient receives a clean colonoscopy report. Since these diagnoses are in between screenings they are considered to be “missed” colorectal cancers and have been associated with higher rates of mortality.
Even though colonoscopies are not radiology-driven, there are several care gaps that AI is well-positioned to close. So what are the AI trends we’re seeing in this space?
Source: American Cancer Society, Inc. Surveillance Research
AI Use-cases in Managing Colorectal Cancer
Similar to breast cancer, the fight against colorectal cancer is largely dependent on adherence to consistent screening. An estimated 60% of colorectal cancer deaths could be prevented with screening, according to observational studies. However, unlike breast cancer, immediate treatments for recognized polyps and abnormalities can occur in real-time during a colonoscopy.
There are four predominant diagnostic tests and screening procedures that gastroenterologists administer to assess colon health and/or detect precancerous abnormalities:
- Stool Tests — the guaiac-based fecal occult blood test (gFOBT) uses a chemical called guaiac to detect blood in the stool. There are also antibody-based tests to detect blood in the stool and tests to determine any altered DNA in the stool.
- Flexible Sigmoidoscopy — the gastroenterologist will insert a short, flexible, lighted tube into your rectum to identify polyps or cancer inside the rectum and lower third of the colon.
- CT Colonography (Virtual Colonoscopy) — a non-invasive, X-ray-based procedure that allows the physician to view digital imaging of the entire colon to detect abnormalities.
- Colonoscopy — the gastroenterologist will insert a short, flexible, lighted tube into your rectum to identify polyps or cancer inside the rectum and the entire colon. This procedure also enables the clinician to remove polyps in real-time and is, therefore, the standard follow-up exam if anything abnormal comes up in prior examinations.
COVID-19 has exacerbated the backlog of patients with missed screening appointments — a product of shifting healthcare resources to address the pandemic and patient aversion towards healthcare settings to minimize personal risk and support healthcare priorities. As health systems begin to refocus their efforts on screening plans, AI can be used to assess the risk level of patients with missed appointments and navigate them to the needed screening procedure that also minimizes the amount of resources lost.
The American Society of Gastroenterologists (ASGE) now recommends that clinicians spend at least 6-10 minutes examining the colon for polyps during the procedure’s withdrawal phase — when they have reached the end of the colon and are wrapping up the procedure. At scale, this extra expense of time to produce a more accurate diagnosis can limit the number of patients a physician can successfully assess. An AI-enabled system similar to Computer-Aided Detection (CAD) for radiology can step in here to notify a physician of perceived polyps, especially in the case of a CT Colonography, helping them improve accuracy and save time.
Furthermore, 25-30% of CRC patients have a family history of the disease. This indicates that CRC can be driven by genetic and shared environmental factors, paving the way for AI to sift through large datasets of genomic data to identify patients that are predisposed to CRC and therefore have more urgent screening needs.
How effective is AI in diagnosing Colorectal cancer?
On the AI front, CT Colonoscopies (Virtual Colonoscopies) are gaining traction in medical environments as a non-invasive alternative to screening practices. CT Colonoscopies make use of X-rays to produce images of the entire colon on a computer screen for the doctor to analyze.
The FDA approved GI Genius in April 2021, a medical device that uses AI to highlight polyps and other dangerous lesions during colonoscopies in real-time. The device has both hardware and AI-enabled software components. As the gastroenterologist performs the colonoscopy, the system produces a low-volume sound and superimposes green markers on the video from the endoscope camera every time it detects a potential polyp to notify the physician of the abnormality.
To assess the safety and effectiveness of the device, the FDA created a randomized, controlled study of 700 subjects who required a colonoscopy for screening purposes, indications from a stool test, or symptoms of colorectal cancer. The main analysis was completed on a subpopulation of 263 patients who had substantial screening experience and data. Around half of the subpopulations had a white light standard colonoscopy with GI genius and the other half did not have GI genius in their colonoscopies.
The study found that colonoscopies that were administered with GI Genius were able to detect lab-confirmed precancerous or cancerous tumors in 55.1% of patients compared to 42.0% of patients with only the baseline colonoscopy. This 13% disparity was clear enough for FDA approval and exemplifies the power that AI has in promoting patient care in the colon health space.
With more innovations like GI Genius seeking regulatory approval, colorectal cancer is yet another prime candidate for AI disruption. We’re already seeing a trend in the practice of gastroenterology for physicians to spend more time analyzing potential polyps. In CRC, AI has been deployed prior to the point of diagnosis to address missed screenings and in real-time during colonoscopies.
AI is ready to pick up the burden of deeper analysis to help gastroenterologists continue what they do best — treating patients.
Shyam Chandra
Shyam is dual-degree student at Northwestern University studying Biomedical Engineering and Communication. He has a passion to leverage the power of data, technology, and entrepreneurship to enhance health outcomes.