Placeholder

3D Surface/Volume Registration for Surgical Simulation

1,000

Product Description

Summary
This project concerns the fields of Image Processing and Computer Vision, and Medical Image
Registration. It aims to extend a system that has been created to register sets of 2D scan data of human
bodies, to a 3D mannequin. This is done in order to create simulated ultrasound for the 3D mannequin,
which can be used to train radiologists in simple procedures such as:
· Drainage of abscess
· Drainage of fluid collection in the abdomen (ascites) and chest (pleural effusion)
· Relief of blockages affecting the kidneys or the liver
· Insertion of catheter tubes into blood vessels
· Biopsy of deep tissues
· Treatment of cancers by radiofrequency ablation (List of applications taken from [11])
First, existing methods and applications in the areas of medical imaging, segmentation and
registration are discussed, and selected for use. Next, discussion is given to the theory behind the
implementation of the proposed extensions to the system. A method is devised to automatically annotate
points to an image in a scan volume. The resulting points are used to align the scan volume to the 3D
mannequin. The project is extended by enabling partial-body scan volumes to be registered to the 3D
mannequin. This is achieved via a pre-processing step where the partial-body scan is registered to a fullbody
scan.
The results of the project are discussed with respect to improvements over the original system’s
manual point annotation and to the accuracy of the registration. The results generated by the extensions
stated in the minimum requirements are favourable, and improve on human annotation, but the partialbody
annotation is not robust and reliable enough to be considered finished. However, it suggests the
basis of a good solution, and can be considered work-in-progress to be developed in the future.
The completed deliverables for this project are:
· Application for automatic physical edge finding and point annotation, and for partial-body scan
volume annotation, with source code
· Written project report, detailing the theory behind the registration process and the image
processing, the implementation of the system, and the results of the new system.

Contents
1 Introduction ……………………………………………………………………………………………………………. 1
1.1 The Problem………………………………………………………………………………………………………. 1
1.2 Minimum Requirements………………………………………………………………………………………. 1
1.3 The Existing System …………………………………………………………………………………………… 3
2 Background and Similar Methods……………………………………………………………………………. 4
2.1 Medical Image Registration …………………………………………………………………………………. 4
2.2 Medical Image Segmentation and Region Identification………………………………………….. 6
2.3 Image Registration ……………………………………………………………………………………………… 6
2.4 General Registration Algorithms ………………………………………………………………………….. 7
2.4.1 Iterative Closest Point (ICP) ………………………………………………………………………….. 7
2.4.2 Random Sample Consensus (RANSAC) …………………………………………………………. 8
3 Solutions …………………………………………………………………………………………………………………. 9
3.1 Solutions for the Initial Problem…………………………………………………………………………… 9
3.1.1 Identifying the Body Region………………………………………………………………………….. 9
3.1.2 Extracting the Contour ………………………………………………………………………………… 12
3.2 Solutions for the Extended Problem ……………………………………………………………………. 14
3.2.1 Aligning the Image Slices ……………………………………………………………………………. 16
3.2.2 Bone Extraction………………………………………………………………………………………….. 16
3.2.3 Sobel edge detection……………………………………………………………………………………. 17
3.2.4 Control Points for Registration …………………………………………………………………….. 19
4 Implementation……………………………………………………………………………………………………… 20
4.1 Methodology ……………………………………………………………………………………………………. 20
4.2 Development Plan …………………………………………………………………………………………….. 21
4.3 Implementing the Solution to the Initial Problem………………………………………………….. 22
4.4 Implementing the Solution to the Extended Problem…………………………………………….. 24
4.4.1 User Interface and priming the algorithm………………………………………………………. 24
4.4.2 Image processing and bone extraction …………………………………………………………… 26
4.4.3 Slice correspondences and point set generation………………………………………………. 26
5 Results ………………………………………………………………………………………………………………….. 27
3D Surface/Volume Registration for Surgical Simulation
IV
5.1 Segmentation and Point Annotation…………………………………………………………………….. 27
5.2 Registration ……………………………………………………………………………………………………… 30
5.3 Partial-body scan registration……………………………………………………………………………… 33
6 Evaluation …………………………………………………………………………………………………………….. 34
6.1 Evaluation of the Minimum Requirements …………………………………………………………… 35
6.1.1 Segmentation and Annotation ………………………………………………………………………. 35
6.1.2 Registration ……………………………………………………………………………………………….. 36
6.1.3 Computer vs. Human Annotation………………………………………………………………….. 36
6.1.4 Robustness…………………………………………………………………………………………………. 38
6.1.5 Improvements…………………………………………………………………………………………….. 39
6.2 Evaluation of the Extended Requirements ……………………………………………………………. 39
6.2.1 Slice Registration and Correspondences ………………………………………………………… 40
6.2.2 Point sets and registration ……………………………………………………………………………. 42
6.2.3 Improvements…………………………………………………………………………………………….. 43
7 Conclusion…………………………………………………………………………………………………………….. 44
References ……………………………………………………………………………………………………………….. 45
Appendix A – Personal Reflection …………………………………………………………………………….. 47
Appendix B – UML ………………………………………………………………………………………………….. 48
B.1 Overview of the VolView system ………………………………………………………………………. 48
B.2 Partial Body Scan Registration process……………………………………………………………….. 49
B.2.1 Initial design/prototype……………………………………………………………………………….. 49
B.2.2 Final design……………………………………………………………………………………………….. 50
Appendix C – Project Plan ……………………………………………………………………………………….. 51
C.1 Initial Project Plan ……………………………………………………………………………………………. 51
C.2 Revised Project Plan…………………………………………………………………………………………. 52
Appendix D – Evaluation Criteria

DOWNLOAD (CHAPTER 1-5)