Automated Assessment of Surgical Quality in Distal Gastrectomy: Development of a Novel Computer Vision Model Based on the Critical View of Quality (CVQ).

OBJECTIVE: To evaluate the clinical validity of the critical view of quality (CVQ) as a measure of lymphadenectomy quality in minimally invasive distal gastrectomy and to develop a computer vision model for automated CVQ assessment. BACKGROUND: Objective intraoperative assessment of lymphadenectomy quality in gastric cancer surgery remains limited, relying largely on postoperative surrogate markers such as lymph node yield. METHODS: This retrospective study included 260 patients who underwent laparoscopic or robotic distal gastrectomy with complete intraoperative video recordings. CVQ was defined as an anatomy-based scoring system reflecting the completeness of lymph node dissection across five stations. Associations between CVQ and lymph node yield were evaluated using correlation and multivariable linear regression analyses. A computer vision model was developed to classify CVQ components as complete or incomplete using temporally contextualized video segments. RESULTS: CVQ demonstrated a moderate positive correlation with lymph node yield (Pearson r=0.485, P<0.001; Spearman ρ=0.484, P<0.001) and remained independently associated after adjustment for clinical covariates (β=4.79, 95% CI 3.83-5.74, P<0.001). Lymph node retrieval increased across CVQ quartiles (28.9 vs. 47.6 nodes, P<0.001). Lower CVQ scores were associated with older age, higher BMI, higher ASA classification, male sex, and surgeon-related variability (all P<0.05). CVQ was not associated with short-term postoperative outcomes. Automated CVQ classification achieved average precision up to 91.5%. CONCLUSIONS: CVQ is a clinically meaningful measure of lymphadenectomy quality that correlates with lymph node yield while capturing operative difficulty and surgeon-related variability. Automated assessment using surgical video is feasible and may enable scalable evaluation of intraoperative performance.