Related Experiment Video
Updated: Apr 18, 2026
Robotized Testing of Camera Positions to Determine Ideal Configuration for Stereo 3D Visualization of Open-Heart Surgery
Published on: August 12, 2021
Development of anatomical metrics for objective quantification of critical view of safety
Benjamin Mueller1, Busisiwe Mlambo2, Sue Kulason2
1Digital Solutions, Intuitive Surgical Inc., 5655 Spalding Drive, Peachtree Corners, GA, 30092, USA. benjamin.mueller@intusurg.com.
Background:
Debates on the subjective criteria for evaluating Surgical Milestones, such as achieving the critical view of safety (CVS) during cholecystectomy, remain a prominent focus and challenge in the field of surgical data science. In this study, we computed anatomical metrics with machine learning tools and investigated the relationship between these objective anatomical metrics and subjective criteria for CVS achievement.
Methods:
We implemented and calibrated a zero-shot monocular depth estimation model for endoscopic images from cholecystectomies. These depth measures were integrated with human-annotated segmentation masks of three key anatomical structures relevant to CVS: cystic duct, cystic artery, and gallbladder. Computational geometry techniques were then employed to extract structure-specific depth distributions and compute two anatomical metrics: diagonal length and surface area. We tested for significant differences in these case-wise metrics, grouped by human-annotated CVS status.
Results:
2256 frames (35 cases) were graded on CVS criteria, of which 343 frames (17 cases) met all three CVS criteria and 384 frames (17 cases) met no CVS criteria. The calibrated depth model achieves 0.063 on absolute relative error and 0.774 on squared relative error in the metric measurement. The diagonal length and surface area of both the cystic duct and cystic artery were significantly larger when all CVS criteria were met. On average, the cystic duct (cystic artery) length was 7.6 mm (13.6 mm) longer when CVS criteria was met.
Conclusion:
In this study, we presented a pipeline that generates anatomical measures from monocular endoscopic images utilizing a depth estimation model, anatomical segmentation masks, and computational geometry. The diagonal length and surface area of the cystic duct and cystic artery were found to be significantly larger in cases where all CVS criteria are met; this lends support to the use of these anatomical metrics as objective grounds for assessing CVS achievement.
