Suturing and knot-tying are foundational skills in surgical education,1 the mastery of which is critical for trainees, as superior techniques improve patient outcomes,2,3 whereas poor performance may lead to increased operation time and complications such as infection, bleeding, etc. Various assessment methods have been developed over the years.4

Traditional assessments of suturing and knot-tying rely on subjective evaluations by experienced surgeons, using tools such as the Objective Structured Assessment of Technical Skills (OSATS)5 and other task-specific checklists.6 While they have sufficient evidence of validity, they are highly subjective, time-intensive, vulnerable to evaluator bias and fail to quantify biomechanical and kinematic skill dimensions.7 These findings highlight the need for more objective and standardized approaches. Currently, artificial intelligence (AI), which is based on predictive modeling methods, has emerged as a promising tool that, seamlessly processes large volumes of data and provides rapid, automated, and reproducible feedback without experts.8

However, AI-based assessments of suturing and knot-tying skills face distinct hurdles. These techniques involve fine motor skills, closely related steps and strict force controls, which require applying precise tension to successfully close the wound, prevent leakage, and avoid damage to the tissue.9 More critically, heterogeneity in the data dimension across surgical contexts severely constrains the generalizability of algorithms. Illustratively, a trajectory curvature analysis model optimized for open surgery may misjudge the “stiff operation” in laparoscopic suturing because of restricted instrument movement. Existing reviews of surgical AI assessment either focus on a single context10, 11, 12, 13 or address shared issues of skill assessment.14,15 While acknowledging the limitation of generalization, they do not provide practical guidance specific to suturing and knot-tying skills. Since these skills are prevalent in open, laparoscopic and robotic surgeries, a specialized review is warranted which not only advises researchers on the data dimension required for model training in different scenarios but also assists educators in ascertaining the reliability of various AI methods for suturing and knot-tying skills assessment.

The aim of this review was to provide practical and updated guidelines for AI-based assessments of suturing and knot-tying skills. By comparing the data input, algorithms, and performance across different models and surgical contexts, we sought to explore the utility of AI methods, identify potential challenges, and propose practical solutions.