Date of Award

11-30-2018

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Science

First Advisor

Dirk Reiners

Abstract

Design and development of a virtual reality based surgical simulation has many steps and the first and most important step is the comprehensive analysis of the surgery. We performed comprehensive analysis, hierarchical task analysis, which allowed steps and goals of the surgery to be understood while expressing the order of execution and hierarchical relations between the tasks of the surgery. Time and performance metrics derived from the comprehensive analysis provides detailed procedural feedback throughout the surgical simulation, which will help classify surgeon’s skill level. We developed quantitative performance metrics for arthroscopy-based rotator cuff surgery with the goal to establish objective skill assessment that can differentiate the performance between novice and expert. Ten shoulder arthroscopic rotator cuff surgeries performed by two novice and fourteen by two expert surgeons were analyzed. Statistical analysis was performed using proposed assessment metrics for each video. Two existing evaluation systems: basic arthroscopic knee skill scoring system (BAKSSS) and the arthroscopic surgical skill evaluation tool (ASSET) were used for validation of our proposed metrics. Virtual surgery simulators allow for physicians to practice with difficult or not common 3D scenarios, but the process of design, creation, and refinement of the 3D models is extensive and laborious. Therefore, we created Generative Anatomy Modeling Language (GAML) for generating variations of 3D virtual human anatomy in real time. The perturbation of the 3D models is satisfied with non-linear geometry constraints to create authentic human anatomy. This integrated non-linear optimization model requires exponential execution time. However, we introduce Partition-based Optimization Model for GAML (POM-GAML), which effectively computes the solution for non-linear optimization model and reduces computation time from exponential to linear time. This is achieved by grouping the 3D geometric constraints into communities with various community detection algorithms (k-means clustering, Clauset Newman Moore, and Density-Based Spatial Clustering of Applications with Noise) and partition the non-linear optimization problem into sub-problems. Our results showed that solving sub-problems of the original model caused computation time to be reduced from exponential time to linear time and the error rate between the partitioned and non-partitioned approach to decrease with the increasing number of constraints.

Download

Share

COinS