NCT07186803

AI and Safety in Laparoscopic Cholecystectomy: A Randomized Controlled Trial

Evaluating the Clinical Impact of Artificial Intelligence on Safety in Laparoscopic Cholecystectomy: A Randomized Controlled Trial

• Status: Recruiting
• Phase: Phase 3
• Study type: Interventional
• Enrollment: 70 (estimated)
• Sponsor: University Health Network, Toronto · Academic / Other
• Sex: All
• Age: 18 Years
• Healthy volunteers: Not accepted

Summary

Today, the majority of gallbladder removals surgeries are done using minimally invasive techniques through small cuts to help patients recover faster. However, these procedures are technically more challenging because surgeons have a restricted view of the patient's anatomy, which can increase the risk of serious complications. Artificial intelligence (AI) tools have been developed to guide surgeons during surgery and help them make safer decisions that reduce the risk of injury to the patient. This study will use a randomized controlled trial to compare outcomes between surgeries with AI assistance and standard procedures without AI. Primary Objective: To determine whether the AI improves surgeons' ability to achieve the Critical View of Safety, a key step for safe gallbladder removal, compared to standard procedures. Secondary Objectives: * Determine whether the AI helps the surgeon perform more safe dissections compared to the standard procedures. * Collect surgeon feedback on the use of AI during the procedure

Detailed description

To measure the clinical impact of artificial intelligence (AI) guidance on the achievement of safety milestones in laparoscopic cholecystectomy compared to standard care, the study team will conduct a randomized controlled trial of 10 surgeons or fellows and 50 patients undergoing laparoscopic cholecystectomy procedures at two hospital sites part of the University Health Network in Toronto, Ontario, Canada (Toronto General Hospital and Toronto Western Hospital). Surgeons or fellows randomized to the intervention group (AI) will each perform 5 procedures using two AI models that provide real-time feedback to guide safe dissections and the achievement of the critical view of safety. Surgeons or fellows randomized to the control group will each perform 5 procedures using the standard care approach. Internal laparoscopic recordings will be collected from both the intervention and control groups for post-operative outcome analysis by blinded expert surgeon reviewers. The research team will evaluate whether the use of AI during the procedure improves the achievement rate of the Critical View of Safety as compared to standard procedures. Additionally, secondary outcomes will be assessed including the proportion of dissections that occurred above the line of safety, surgeon feedback on the use of AI during the procedure, observational notes recorded by the research coordinator present during each procedure, and 30-day post operation chart review.

Conditions

• Laparoscopic Cholecystectomy

Interventions

Type	Name	Description
DEVICE	Artificial Intelligence Guidance Models	The intervention will involve the use of two artificial intelligence (AI) models to provide surgical guidance during laparoscopic cholecystectomy procedures. The AI models will provide real-time feedback based on the live surgical feed (internal patient anatomy captured by laparoscopic camera) displayed on an operating room monitor. The GoNoGoNet model identifies safe and unsafe zones of dissection. This is done by showcasing a green overlay over safe zones of dissection, and a red overlay over unsafe zones of dissection. The DeepCVS model provides text-based feedback based on its assessment of the following three criteria defining the Critical View of Safety: 1) complete clearance of the hepatocystic triangle from fat and fibrous tissue, 2) only two structures visible entering the gallbladder (cystic artery and duct) and 3) the lower third of the gallbladder must be dissected off the liver bed, exposing the cystic plate.

Timeline

• Start date: 2025-09-17
• Primary completion: 2026-06-30
• Completion: 2026-07-30
• First posted: 2025-09-22
• Last updated: 2026-01-13

Locations

2 sites across 1 country: Canada