If the surgeon is trained, they usually need supervision by an experienced doctor who can teach them the technique. That may be changing with a new artificial intelligence system developed by Caltech researchers and USC urologist her Keck Medicine, which aims to give surgeons valuable feedback on the quality of their work. there is.
The goal of the new Surgical AI System (SAIS) is to provide surgeons with objective performance assessments to improve their work and thus improve patient outcomes. When a surgical video is provided, SAIS can identify the type of surgery being performed and the quality of the surgery performed by the surgeon.
The system was introduced through a series of articles published simultaneously in the journals Nature Biomedical Engineering, npj Digital Medicine, and Communications Medicine at the end of March 2023.
“In high-risk environments such as robotic surgery, it is not realistic for AI to replace human surgeons in the short term,” said Anima Annan, Bren Professor of Computing and Mathematical Sciences and senior author of the study. Kumar says. “Instead, we are focused on making human surgeons better and more effective through AI, as we looked at how AI can safely improve surgical outcomes for patients.”
SAIS was trained using large amounts of video data annotated by medical experts. Surgeons’ performance was assessed to the level of individual individual movements. That is, I held the needle, stuck it into the tissue, and pulled it out of the tissue. After training, SAIS was tasked with reviewing and evaluating the surgeon’s performance in various procedures using videos from various hospitals.
“SAIS has the potential to provide accurate, consistent, and scalable feedback from surgeons,” said lead author of the study, a former postdoctoral fellow at Caltech and now a senior AI engineer at Vicarious Surgical. said Dani Kiyasseh, Researchers say SAIS is expected to provide guidance to surgeons on what skill sets they need to improve.
To make this tool more valuable to surgeons, the team developed AI capabilities to justify skill assessment. AI is now able to inform surgeons of their skill level and point them to specific video clips to provide detailed feedback on the basis for their assessment.
“We were able to show that AI-based explanations like this often match what a surgeon would provide,” says Kiyasseh. “Trusted AI-based explanations can pave the way for providing feedback when fellow surgeons are not immediately available.”
Researchers who tested SAIS early on found that unintended bias crept into the system, allowing the AI to assess whether it was more or less proficient than experience would suggest based solely on an analysis of the surgeon’s gross movements. I just realized something. To address this issue, the researchers guided his AI system to focus only on relevant aspects of the surgical video. Narrowing the focus didn’t eliminate the bias, but it reduced it.
“Human-derived surgical feedback is currently neither objective nor scalable,” says Andrew Hung, urologist at USC’s Keck Medicine and associate professor of urology at USC’s Keck School of Medicine. “AI-derived feedback, such as that provided by our system, offers great opportunities to provide actionable feedback to surgeons.”
These studies are: “Vision Transformers for Decoding Surgeon Activities from Surgical Videos”, “Human Visual Explanations Reduce Biases in AI-Based Surgeon Skill Assessment”, “Using Artificial Intelligence to A multicenter study providing reliable and unbiased feedback to patients”. surgeon. The study was funded by the National Cancer Institute.
