AI Blood-Pixel Measure Associated With Laparoscopic Liver Resection Skill Scores

A retrospective analysis of 146 laparoscopic liver resection videos assessed whether an AI-derived blood-pixel count tracked surgical proficiency scores. The measure was associated with lower scores but had limited sensitivity, and the authors call for external validation.

Retrospective analysis included 146 submitted surgical videos.
Blood-pixel counts were associated with lower proficiency scores.
At 90.2% specificity, sensitivity for low scores was 29.2%.
External validation and prospective studies are needed.

Researchers retrospectively analyzed 146 laparoscopic liver resection videos submitted to Japan’s Endoscopic Surgical Skill Qualification System in 2020–2021. A DeepLabv3+ model segmented bleeding regions, and investigators calculated the total number of pixels classified as blood across sampled frames. They compared this blood-pixel count (BPC) with qualification scores, using a 70-point cutoff for pass/fail and defining low scores as those at least one standard deviation below the mean.

Median BPC was 1,741,749 pixels. Counts increased across proficiency groups (P<0.001), and failed videos had higher BPC than passed videos (P=0.001). BPC discriminated low-score videos with an area under the ROC curve (AUC) of 0.708 and failed videos with an AUC of 0.648. At a threshold selected for 90% specificity, sensitivity for identifying low scores was 29.2%. In low-difficulty cases, the AUC for low-score identification was 0.784.

The findings support BPC as a potential high-specificity flag for expert review, not as a stand-alone assessment of competence. The study was retrospective and limited to videos submitted to one qualification system; external validation and prospective implementation are needed before its use in assessment or feedback systems can be established.

In practice

For teams assessing laparoscopic liver resection videos, AI-derived blood-pixel count was associated with qualification scores and may help flag some videos for expert review. However, at the selected high-specificity threshold, sensitivity for low scores was only 29.2%, so many low-scoring videos would not be identified. This retrospective analysis covered 146 videos submitted to Japan’s qualification system; external validation and prospective implementation studies are needed. The findings do not establish BPC as a stand-alone measure of surgical competence.

AI-generated clinical interpretation · Not editor-reviewed

AI summary · Not yet editor-reviewed

Is this summary clinically accurate?

Help fellow clinicians: your rating sends inaccurate summaries straight to our editors.

Source

Surgical Endoscopy: Proof-of-concept study toward the automated evaluation of surgical skills in laparoscopic liver resection using artificial intelligence ↗

This is an automated AI-condensed summary that has not yet been reviewed by an editor. Always consult the full item at the original source.