< + > The AI That Watches You Operate: Inside Stanford’s Breakthrough for Surgical Residents

I am always looking for unique, high-impact uses of AI in healthcare. While attending the annual Society of American Gastrointestinal and Endoscopic Surgeons conference (SAGES), I sat through a presentation that completely delivered by modernizing the surprisingly manual world of surgical training.

Dr. Chloe Nobuhara, a Stanford general surgery resident, showcased a tool originally conceived by her Principal Investigators, Dr. Yeung-Levy and Dr. Jeff K. Jopling. What they built is an AI application that watches laparoscopic surgical video, analyzes the surgeon’s performance, and gives them the feedback they need to improve.

Here is the inside scoop on how this innovation works.

Key Takeaways from the Discussion with Dr. Nobuhara at SAGES 2026

• The Time Trap. Teaching hospitals are drowning in unwatched surgical video sitting on thumb drives. By running this footage through a specialized AI (dubbed Surgical Learning Model or SLM) to automatically segment the operation and analyze technique, residents receive instant, targeted feedback without draining an attending physician’s schedule.
• Generic AI Fails in the OR. Off-the-shelf LLMs choke on the length and context of multi-hour surgical footage. Training a dedicated model on specific anatomical and directional language creates a highly accurate tool that understands the nuances of the operating room.
• The Limits of Manual Assessment. Relying solely on humans to assess surgical skill is inherently subjective and difficult to scale. Establishing an AI-driven, objective baseline for surgical performance today paves the way for fairer board certifications and prepares the industry for the impending era of autonomous robotics.

Solving the “Thumb Drive” Dilemma

What I didn’t realize and what Dr. Nobuhara highlighted, was how manual and time-consuming surgical training really is. Most laparoscopic surgeries by surgeons in training are recorded, but there is not enough the time to review them.

“We as residents and attendings have hundreds of hours of video that we’ve collected,” explained Dr. Nobuhara. “We don’t go through it nearly as often as we should because it’s a time-consuming process. In the ideal world, it’s an attending and a resident sitting down reviewing the case that they just did together.”

To fix this, Dr. Yeung-Levy, Dr Jopling, Dr. Nobuhara and their team built computer vision models into a web app. Residents upload their videos, and the AI goes to work detecting errors and assessing the skill of the surgeon.

“The idea here is that residents can self-study on their own,” continued Dr. Nobuhara. “They can watch the videos, get an AI-generated summary, and then come to the attending the next day or when they have time to review together.

Even better, the AI automatically “chunks” the operation into distinct segments. For example, with a typical laparoscopic cholecystectomy operation, their app divides it into five steps. This allows educators to instantly pull specific clips of basic “clipping and cutting” for interns, or complex dissections for senior residents. It transforms a lengthy three-hour video file into a precise moments that can help viewers focus on specific points in time – a fantastic teaching tool.

Building a Surgical Learning Model (SLM)

So, why not just upload these videos to Gemini or Claude? Because off-the-shelf LLMs today choke on the length and context of a multi-hour operation.

“We built out a specialized surgical video language model that can use more specific surgeon language, including anatomy,” explained Dr. Nobuhara.

I called this a “Surgical Learning Model” (SLM) and it is already being piloted by over 60 users across four hospitals. The feedback has been overwhelmingly positive – users trust the AI’s segmentation and love the efficiency it offers.

The Stanford team is already looking beyond the resident’s video review. They are laying the groundwork for broader applications, like objective board certifications and the coming age of autonomous robotics.

“How do we grade the robots that are doing surgeries on their own?” is a question that Dr. Nobuhara and the team at Stanford are now asking.

The Bottom Line

AI might not be replacing surgeons anytime soon, but tools like this can help them improve their skills. In the end, better-trained surgeons mean safer patients which is a future worth getting behind.

Learn more about Standford at https://med.stanford.edu/

Listen and subscribe to the Healthcare IT Today Interviews Podcast to hear all the latest insights from experts in healthcare IT.

And for an exclusive look at our top stories, subscribe to our newsletter and YouTube.

Tell us what you think. Contact us here or on Twitter at @hcitoday. And if you’re interested in advertising with us, check out our various advertising packages and request our Media Kit.