All in on AI

Products both currently available and in development use artificial intelligence to advance GI care.

Harvard sociologist Daniel Bell once said, “Technology, like art, is a soaring exercise of the human imagination.” We should keep Mr. Bell’s thought in mind as we consider how artificial intelligence (AI) is revolutionizing all outpatient surgery specialties. Gastroenterology is no exception.

As someone who has witnessed the birth and subsequent rise of this industry over the past decade, I marvel at the rapid advances AI is making in surgical care — and the possibilities that lie ahead. Diagnostic precision has improved, efficiency has increased exponentially and margins of error continue to shrink thanks to AI assistance.

Adoption awaits

As an ASC owner, it’s important to focus on modalities that can be incorporated into outpatient settings. Between current technologies and innovations on the horizon, major players in the field are bringing more products to the table.

Of note, FDA approval exists for some, but not all, of these AI-based GI technologies. The regulatory landscape is evolving, and ongoing clinical validation and regulatory review is underway for broader adoption of these technologies in the U.S.

From AI-assisted colonoscopies detecting polyps with increasing accuracy to smart capsule endoscopies with unprecedented diagnostic capabilities, these advancements are streamlining procedures, improving patient outcomes, reducing costs and transforming the modern outpatient GI landscape.

• AI-assisted colonoscopy systems all use convolutional neural networks (CNNs) to detect polyps and highlight adenomas in real time. As an early adopter of this technology, I can attest to the increased efficiency and clinical value these tools bring to the GI suite. They are especially beneficial in training environments for young fellows.

A 2020 study published in Gastroenterology reported that one platform from a major manufacturer improved the adenoma detection rate (ADR) by 14%. Another showed a nearly 50% decrease in missed polyps compared to standard colonoscopy. This technology will get even more sophisticated as it matures.

• Smart capsule endoscopy has been around for years. I remember tediously going through hours of video capsule endoscopy images as a young fellow. AI enhancements have taken these technologies to new heights. One AI-assisted system generates more than 50,000 images during a single study, and the time it takes to review all those images has been reduced by up to 40%. CNN-based modules are also being developed to further automate interpretation.

These improvements have significantly enhanced diagnostic accuracy in detecting small intestinal conditions such as Crohn’s disease, obscure gastrointestinal bleeding and tumors — with sensitivity rates reported between 92% to 96%. For outpatient centers without access to deep enteroscopy, AI-enhanced capsule endoscopy offers a powerful, non-invasive alternative. We continue to fight for insurance coverage of these tests.

• Computer-aided histology diagnosis (CADx) in polyp characterization systems enable in vivo histologic prediction with up to 94% accuracy in distinguishing hyperplastic from adenomatous polyps. These tools use advanced imaging technologies such as narrow-band imaging and endocytoscopy to support real-time decision-making during colonoscopies.

A 2021 study found that one product achieved a negative predictive value of 91% for adenomatous histology, while another platform demonstrated 96.9% accuracy in adenoma detection. These systems support “diagnose-and-leave” and “resect-and-discard” strategies, reducing pathology costs, avoiding unnecessary resections and improving procedural efficiency.

• AI-enhanced Barrett’s esophagus surveillance platforms use AI to significantly improve the detection of Barrett’s esophagus and high-grade dysplasia. Studies suggest an 83% increase in Barrett’s detection and an 88% improvement in identifying high-grade dysplasia.

We have always relied on endoscopic visuals and pathological sampling for Barrett’s esophagus. AI now augments this process by leveraging tools to identify early neoplastic changes with greater precision. While I wouldn’t call this a significant improvement over conventional techniques, current estimates suggest AI-assisted systems operate in the 80% to 90% accuracy range — a major step forward in esophageal cancer surveillance.

• AI-based colonoscopy quality monitoring products track key indicators during colonoscopies, including withdrawal time, mucosal coverage and blind spot detection. These parameters are vital for effective polyp detection. For example, maintaining a withdrawal time of at least six minutes is associated with higher ADRs.

These AI systems assist endoscopists during live procedures and are also valuable educational tools for GI fellows by reinforcing performance standards and enabling structured feedback.

• Real-time GI bleeding detection systems are capable of identifying active gastrointestinal bleeding — whether from angiodysplasias, ulcers or neoplasms — with a sensitivity of 89% to 94%. These systems hold significant potential for outpatient management of stable GI bleeding cases, enabling prompt triage and intervention.

Beyond detection, these modalities can support procedural planning and help physicians anticipate therapeutic steps in real time. Think of these as a giant ancillary support-hand for the very complex field of occult GI bleeding.

• AI-guided ERCP navigation and stone detection, while more relevant to inpatient settings, mark another technical advance in GI treatment. These systems use deep reinforcement learning and computer vision to assist with ductal navigation during endoscopic retrograde cholangiopancreatography (ERCP) and to identify bile duct stones. A 2022 pilot study for one platform demonstrated a 93% sensitivity in the detection of common bile duct stones in real time.

It could be especially beneficial in high-volume outpatient centers that offer ERCP, where quick cannulation reduces procedure time and complication risks.

• AI-driven H. pylori detection via breath test analysis uses machine learning algorithms to interpret volatile organic compounds in exhaled breath for non-invasive H. pylori detection. A clinical study of 2,000 patients in Europe showed one product achieved 92% sensitivity and 90% specificity compared to urea breath tests and biopsy.

I see a clear role for this modality in outpatient GI clinics screening for H. pylori without invasive endoscopy. H. pylori testing continues to be a contentious billing issue with private payors, however.

A ‘PILL’ FOR THAT Smart capsule technologies are hitting the market to better detect intestinal conditions.

• AI-optimized bowel prep monitoring tools are being used to assess the adequacy of bowel preparation prior to colonoscopy, a critical factor for procedural accuracy. These AI systems have demonstrated up to 88% accuracy, as supported by peer-reviewed studies. By identifying suboptimal preps in advance, they help optimize scheduling, reduce repeat procedures and improve detection rates.

• Automated video annotation and report generation platforms automate the labeling of anatomic landmarks, detection of polyps and even suggest CPT coding. These tools can save endoscopists several minutes per case on average. Published literature confirms that AI systems reduce documentation time by approximately eight to 10 minutes per case. The result is enhanced workflow efficiency and improved billing accuracy in outpatient settings.

• Autonomous polyp measurement tools from several manufacturers are developing AI platforms for inclusion in the tools that would be capable of measuring polyp size with a precision of up to 1.5 mm. While still evolving, these tools hold promise for standardizing polyp measurement, which is an important factor to determine surveillance intervals and therapeutic decisions.

• AI-powered GI training simulators use AI to improve endoscopy training through real-time performance feedback and simulated scenarios. Studies have shown that trainees using these simulators demonstrate a 30% to 40% improvement in efficiency and technical skill acquisition. These platforms are particularly valuable in outpatient centers that serve as training hubs.

• AI-enhanced digital pathology platforms leverage AI to detect dysplasia in colorectal biopsy samples with a level of precision that surpasses conventional microscopy. These systems reduce diagnostic errors and shorten turnaround times, providing a clear advantage to outpatient GI pathology labs seeking speed and accuracy.

• Wireless motility and pH monitoring devices with AI offer enhanced analyses of gastrointestinal transit times and pH levels. Compared to traditional diagnostic tools, one such product has shown a significant improvement in diagnostic yield for conditions like gastroparesis and slow-transit constipation. These devices are essential additions to the outpatient evaluation of functional GI disorders.

Happening now

At this point in time, it’s clear we’re just scratching the surface of what AI can do for outpatient GI care. What excites me most is that all these advances aren’t forecast for the distant future — they’re happening now.

I was trained by the generation that peered into an endoscope’s handle while the nurse guided the scope. It’s incredible how far we have come in just a few decades. In just a few more years, many of these new advances will already be considered old technologies. That speaks volumes about the speed and force of the AI movement.

For those running ambulatory surgery centers, this is more than innovation — it’s a call to evolve. Personally, I’m all in. OSM