Adaptive Neuro-fuzzy Inference System for Automated Skill Assessment in Robot-Assisted Minimally Invasive Surgery

The new kinds of minimally invasive surgical tools and methods require new and complex skillsets from the surgeons, thus the objective skill-based training became essential. This new domain can greatly rely on the automated assessment of surgical skills. There are many devices and approaches to measure different aspects of surgical task accution, however, the results are often hard to interpret. This paper describes an adaptive neuro-fuzzy inference system for the classification of the performance of subjects using a skill assessment device designed for psychomotor skill training for robotic surgical procedures (the FRS Dome). The FRS Dome offers 7 independent surgical tasks, this paper focuses on two of them, since the rest were often impossible to perform for novices. The fuzzy-systems for the two tasks were designed based on 27 performances of subjects with varying skill-levels, the system is capable of scoring the tasks separately, and also scoring the whole performance on all the tasks together on 1-3 scales. In general, the neurofuzzy system is capable of optimizing the classification based on future measurements, thus our method will be able to tune the classification of the rest of the tasks based on upcoming trial results.