Data-driven assessment of technical skills in minimally invasive surgery

This thesis explores the integration of data-driven assessment into minimally invasive surgical (MIS) training. By embedding force-, motion-, and time-based metrics into simulation curricula, it aims to provide objective, individualized feedback and enable competency-based progression. Across the studies, more than 50,000 performances were analyzed, making this one of the largest datasets on MIS skill development. Part I (Chapter 2) reviews 35 studies on data-driven tools for laparoscopic training. Although validity evidence has steadily grown, implementation into curricula remains inconsistent. Metrics and targeted skills varied widely, with most systems focused on motion analysis while domains such as tissue handling and safe force application were underrepresented. Part II examines the application of data-driven assessment in laparoscopic training. A pilot study (Chapter 3) showed that the Lapron box simulator could differentiate novices from experts and support autonomous at-home training, with all participants reaching proficiency. A multicenter prospective cohort (Chapter 4) used regression-based learning curves to visualize progress, identifying both underperforming residents and those ready for advancement, laying the foundation for individualized training. Chapter 5 compared assessment methods for progression from basic to advanced suturing, finding good skill transfer but highlighting the need to align tools with training goals. Chapter 6 showed that haptic exploration before training improved tissue handling, supporting inclusion of sensory elements. Prediction and retention were addressed in Chapters 7 and 8. From early performance data, the required number of repetitions to reach proficiency could be predicted after only three attempts, enabling personalized schedules. A retention study showed significant deterioration after four months, emphasizing the need for maintenance training. Part III extends these principles to advanced and robotic MIS. Chapter 9 introduced a TaTME training platform with force and torque sensors. Skills improved, but excessive force remained common, stressing the need for repetitive training with feedback. Chapter 10 surveyed European surgeons on steerable laparoscopic instruments (SLI). While interest was high, routine use was limited by concerns about learning curves, safety, cost, and sustainability. Chapter 11 presented a reusable, modular SLI prototype that combined robust mechanics with improved reusability. Robotic training was the focus of Chapters 12–15. Chapter 12 validated force and motion metrics for robotic tissue handling. Chapter 13 showed that experts transferred skills seamlessly between laparoscopy and robotics, while novices and intermediates did not, underlining the need for structured training. Chapter 14, using ex vivo cholecystectomy, demonstrated that robotic training led to higher OSATS scores, fewer errors, and greater motion efficiency, with participants also reporting better ergonomics. Chapter 15 introduced PoLaRS, a portable, robot-independent VR simulator, which showed comparable outcomes to the da Vinci Skills Simulator while offering affordability and accessibility. General discussion This thesis highlights the transformative role of data-driven assessment in MIS education. Validated metrics enable real-time feedback, detailed learning curve analysis, and early detection of underperforming trainees. Compared with traditional methods, this approach supports deliberate practice, peer benchmarking, and flexible, remote training. Artificial intelligence and machine learning promise further advances, such as adaptive feedback and predictive modeling. Ultimately, data-driven, competency-based curricula provide a scalable and personalized framework for surgical training, improving skill acquisition, standardization, and patient safety.