The Next Decade of AI in Surgical Training: From Basic Metrics to Better Surgeons

The Assessment Bottleneck

The current limitation in medical education is the heavy reliance on subjective human observation to evaluate trainee surgeons. Senior doctors simply do not have the hours required to watch every movement a resident makes in the operating theatre. A comprehensive review of AI in surgical training suggests this technology is the exact tool needed to break this assessment bottleneck.

The Push for Objective Evaluation

For decades, surgical proficiency has been judged by human raters standing over a trainee's shoulder. This traditional process is inherently difficult to scale and prone to natural human bias. To improve how surgical performance is assessed, training programmes need a faster, more reliable way to ensure new surgeons are developing the right techniques.

Relying solely on senior staff to grade every practice suture is difficult to scale effectively. The medical field requires a method to evaluate skill that improves the objectivity and scalability of surgical education without draining the limited time of experienced consultants.

How AI in Surgical Training Works Today

Researchers analysed medical literature published between 2015 and 2025 to track how machine learning evaluates surgical performance. While current evidence is largely based on controlled study environments across specific surgical disciplines rather than widespread clinical deployment, the review found that artificial intelligence successfully automates skill assessment.

The systems measure performance through several specific methods:

• Computer vision to track instrument movements and positioning.
• Kinematics to measure the physical forces applied during a procedure.
• Gesture analysis to evaluate overall hand-eye coordination.

Currently, these models excel at binary classification, reliably distinguishing a novice from an expert. The AI ratings show strong agreement with scores given by human experts, confirming the software can accurately measure basic competence. Furthermore, the technology provides automated feedback that has actively improved surgeon performance metrics, particularly for those who were initially underperforming in their cohorts.

The Next Ten Years of the Operating Theatre

While the current models are adept at detecting large differences in skill, the next five to ten years will likely see a shift toward highly specific, algorithmic coaching. The review notes that today's feedback is relatively rudimentary, often just telling a trainee if they passed or failed a motion. However, the trajectory points toward systems capable of doing more fine-tuned skill assessments and generating detailed, constructive feedback.

If these models continue to improve, they could fundamentally shift how we approach surgical proficiency. Rather than relying on broad, subjective evaluations, trainees could receive highly detailed feedback tailored to their specific technical shortcomings. This level of scalable, objective instruction could drastically improve the efficiency of surgical education.

The review also measured how AI categorises live feedback from human instructors during real operations. This suggests future models could train senior surgeons on how to deliver better verbal instructions to their juniors. By analysing which types of human feedback lead to the best trainee outcomes, the AI could help optimise the entire teaching process.

Over the next decade, machine learning will likely transition from a simple grading tool into a continuous, objective co-pilot for medical education. The ability to practice complex procedures with an untiring, automated tutor could fundamentally alter how we assess and scale surgical skills in the future.