Use this URL to cite or link to this record in EThOS:
Title: Segmentation and deformable modelling techniques for a virtual reality surgical simulator in hepatic oncology
Author: Chi, Ying
ISNI:       0000 0004 2676 3241
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 2009
Availability of Full Text:
Access from EThOS:
Access from Institution:
Liver surgical resection is one of the most frequently used curative therapies. However, resectability is problematic. There is a need for a computer-assisted surgical planning and simulation system which can accurately and efficiently simulate the liver, vessels and tumours in actual patients. The present project describes the development of these core segmentation and deformable modelling techniques. For precise detection of irregularly shaped areas with indistinct boundaries, the segmentation incorporated active contours - gradient vector flow (GVF) snakes and level sets. To improve efficiency, a chessboard distance transform was used to replace part of the GVF effort. To automatically initialize the liver volume detection process, a rotating template was introduced to locate the starting slice. For shape maintenance during the segmentation process, a simplified object shape learning step was introduced to avoid occasional significant errors. Skeletonization with fuzzy connectedness was used for vessel segmentation. To achieve real-time interactivity, the deformation regime of this system was based on a single-organ mass-spring system (MSS), which introduced an on-the-fly local mesh refinement to raise the deformation accuracy and the mesh control quality. This method was now extended to a multiple soft-tissue constraint system, by supplementing it with an adaptive constraint mesh generation. A mesh quality measure was tailored based on a wide comparison of classic measures. Adjustable feature and parameter settings were thus provided, to make tissues of interest distinct from adjacent structures, keeping the mesh suitable for on-line topological transformation and deformation. More than 20 actual patient CT and 2 magnetic resonance imaging (MRI) liver datasets were tested to evaluate the performance of the segmentation method. Instrument manipulations of probing, grasping, and simple cutting were successfully simulated on deformable constraint liver tissue models. This project was implemented in conjunction with the Division of Surgery, Hammersmith Hospital, London; the preliminary reality effect was judged satisfactory by the consultant hepatic surgeon.
Supervisor: Cashman, Peter ; Kitney, Richard Sponsor: Lee Family Scholarship
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral