Advances in automated surgery skills evaluation
Abstract
Training a surgeon to be skilled and competent to perform a given surgical procedure, is an important step in providing a high quality of care and reducing the risk of complications. Traditional surgical training is carried out by expert surgeons who observe and assess the trainees directly during a given procedure. However, these traditional training methods are time-consuming, subjective, costly, and do not offer an overall surgical expertise evaluation criterion. The solution for these subjective evaluation methods is a sensor-based methodology able to objectively assess the surgeon's skill level. The development and advances in sensor technologies enable capturing and studying the information obtained from complex surgery procedures. If the surgical activities that occur during a procedure are captured using a set of sensors, then the skill evaluation methodology can be defined as a motion and time series analysis problem. This work aims at developing machine learning approaches for automated surgical skill assessment based on hand motion analysis. Specifically, this work presents several contributions to the field of objective surgical techniques using multi-dimensional time series, such as 1) introduce a new distance measure for the surgical activities based on the alignment of two multi-dimensional time series, 2) develop an automated classification framework to identify the surgeon proficiency level using wrist worn sensors, 3) develop a classification technique to identify elementary surgical tasks: suturing, needle passing, and knot tying , 4) introduce a new surgemes mean feature reduction technique which help improve the machine learning algorithms, 5) develop a framework for surgical gesture classification by employing the mean feature reduction method, 6) design an unsupervised method to identify the surgemes in a given procedure.
Degree
Ph. D.