Skills Assessment in Arthroscopic Surgery by Processing Kinematic, Force, and Bio-signal Data

Arthroscopic surgery is a type of Minimally Invasive Surgery (MIS) performed in human joints, which can be used for diagnostic or treatment purposes. The nature of this type of surgery makes it such that surgeons require extensive training to become experts at performing surgical tasks in tight environments and with reduced force feedback. MIS increases the possibility of erroneous actions, which could result in injury to the patient. Many of these injuries can be prevented by implementing appropriate training and skills assessment methods. Various performance methods, including Global Rating Scales and technical measures, have been proposed in the literature. However, there is still a need to further improve the accuracy of surgical skills assessment and improve its ability to distinguish fine variations in surgical proficiency. The main goal of this thesis is to enhance surgical, and specifically, arthroscopic skills assessment. The optimal assessment method should be objective, distinguish between subjects with different levels of expertise, and be computationally efficient. This thesis proposes a new method of investigating surgical skills by introducing energy expenditure metrics. To this end, two main approaches are pursued: 1) evaluating the kinematics of instrument motion, and 2) exploring the muscle activity of trainees. Mechanical energy expenditure and work are investigated for a variety of laparoscopic and arthroscopic tasks. The results obtained in this thesis demonstrate that expert surgeons expend less energy than novice trainees. The different forms of mechanical energy expenditure were combined through optimization methods and machine learning algorithms. An optimum two-step optimization method for classifying trainees into detailed levels of expertise is proposed that demonstrates an enhanced ability to determine the level of expertise of trainees compared to other published methods. Furthermore, performance metrics are proposed based on electromyography signals of the forearm muscles, which are recorded using a wearable device. These results also demonstrate that the metrics defined based on muscle activity can be used for arthroscopic skills assessment. The energy-based metrics and the muscle activity metrics demonstrated the ability to identify levels of expertise, with accuracy levels as high as 95% and 100%, respectively. The primary contribution of this thesis is the development of novel metrics and assessment methods based on energy expenditure and muscle activity. The methods presented advance our knowledge of the characteristics of dexterous performance and add another perspective to quantifying surgical proficiency