This thesis is aimed at developing a novel method for bone-surface registration with respect to a robot coordinate system when used in orthopedic medical interventions. The objective of the proposed work is to demonstrate that bone surface can be registered using two cameras and a light-analysis scheme. The motivation for the method, its implications, and its potential implementation are discussed in this document.

Although the registration method proposed here is aimed at identifying bone in general, the special case of pedicle-screw placement using robots is addressed as it is one of the most important interventions where robots are poised to make a significant contribution. This case will serve as a verifying test-bed for the proposed registration system.

TABLE OF CONTENTS

1. Itroduction *

2. Literature Review *

2.1 Motivation for Pedicle Screw Insertion
2.2 3D Vision Systems
2.3 Model Matching
2.4 Robotic-Assisted Surgery and Pedicle-Screw Insertion Attempts 3. Image Acquisition * 3.1 The Image Acquisition System
3.2 Stereo Systems
3.3 Camera Calibration
3.4 Robust Calibration 4. Color Correspondence * 4.1 Setup
4.2 Illumination
4.3 Color Sensing
4.4 Algorithm
4.5 Efficiency and Challenges 5. Model Matching * 5.1 Difficulties
5.2 Conceptual Modeling
5.3 Model Implementation
5.4 Search Space
5.5 Search Mode 6. Accuracy Analysis and System Validation * 6.1 Acquisition Error
6.2 Matching Error
6.3 Validation Steps 7. CONCLUSIONS AND RECOMMENDATIONS *

REFERENCES *