Date of Award

Fall 11-22-2024

Document Type

Dissertation

Degree Name

Ph.D. in Mechanical Engineering

Organizational Unit

Daniel Felix Ritchie School of Engineering and Computer Science, Mechanical and Materials Engineering

First Advisor

Paul Rullkoetter

Second Advisor

Casey Myers

Third Advisor

Mohammad Mahoor

Fourth Advisor

Xin Fan

Copyright Statement / License for Reuse

All Rights Reserved
All Rights Reserved.

Keywords

Machine learning, Computer vision systems, Pose estimation, Surgical navigation

Abstract

Machine learning has emerged as a key technology for enabling advanced computer vision systems. These systems now permeate many industries, and have enhanced traditional processes through autonomous interpretation of visual data. In the field of orthopaedics, the increasing prevalence of imaging highlights a need for similar tools. In many cases, however, direct translation of available frameworks fails to resolve the complex problems currently facing this field. Stringent performance requirements, complex visual environments, data scarcity, and implications for patient safety represent domain-specific factors which pose unique challenges to proven techniques. Novel approaches are needed to realize the full benefits of visual understanding from orthopaedic modalities.

This dissertation contributes to expanding the role of computer vision in the field of orthopaedics. Progress is made along two distinct axes spanning both research and industry settings. The first part of this work concerns 6 degree-of-freedom pose estimation of native anatomy in dynamic stereo-radiography. Frameworks are introduced which can localize skeletal structures of the knee and shoulder with a high degree of accuracy, and with no required manual interaction. The resulting systems may catalyze large-scale investigative efforts or clinical use cases through efficient measurement of in vivo movement. The second part of this work focuses on the role of computer vision in orthopaedic procedures. It is contended that markerless computer vision algorithms based on machine learning may benefit intraoperative monitoring and navigation by alleviating limitations of existing technical approaches. Two applications are then investigated: pose estimation of a surgical instrument, and image-guided surgical navigation for robot-assisted drilling. In the first application, the accuracies of pose estimates fail to meet requirements associated with navigation systems. Recommendations for future improvements are thus presented. In the second project, the realized navigation framework is shown to replicate preoperative plans with precision that is on par with commercial robotic systems. The experiments and methods described in this work represent small but significant steps towards harnessing computer vision to improve patient outcomes in orthopaedics.

Copyright Date

11-2024

Publication Statement

Copyright is held by the author. User is responsible for all copyright compliance.

Rights Holder

William Stewart Burton II

Provenance

Received from author

File Format

application/pdf

Language

English (eng)

Extent

259 pgs

File Size

161 MB

Download



Share

COinS