Use this URL to cite or link to this record in EThOS:
Title: Registration of magnetic resonance and ultrasound images for guiding prostate cancer interventions
Author: Hu, Y.
Awarding Body: University College London (University of London)
Current Institution: University College London (University of London)
Date of Award: 2013
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Prostate cancer is a major international health problem with a large and rising incidence in many parts of the world. Transrectal ultrasound (TRUS) imaging is used routinely to guide surgical procedures, such as needle biopsy and a number of minimally-invasive therapies, but its limited ability to visualise prostate cancer is widely recognised. Magnetic resonance (MR) imaging techniques, on the other hand, have recently been developed that can provide clinically useful diagnostic information. Registration (or alignment) of MR and TRUS images during TRUS-guided surgical interventions potentially provides a cost-effective approach to augment TRUS images with clinically useful, MR-derived information (for example, tumour location, shape and size). This thesis describes a deformable image registration framework that enables automatic and/or semi-automatic alignment of MR and 3D TRUS images of the prostate gland. The method combines two technical developments in the field: First, a method for constructing patient-specific statistical shape models of prostate motion/deformation, based on learning from finite element simulations of gland motion using geometric data from a preoperative MR image, is proposed. Second, a novel “model-to-image” registration framework is developed to register this statistical shape model automatically to an intraoperative TRUS image. This registration approach is implemented using a novel model-to-image vector alignment (MIVA) algorithm, which maximises the likelihood of a particular instance of a statistical shape model given a voxel-intensity-based feature vector that represents an estimate of the surface normal vectors at the boundary of the organ in question. Using real patient data, the MR-TRUS registration accuracy of the new algorithm is validated using intra-prostatic anatomical landmarks. A rigorous and extensive validation analysis is also provided for assessing the image registration experiments. The final target registration error after performing 100 MR–TRUS registrations for each patient have a median of 2.40 mm, meaning that over 93% registrations may successfully hit the target representing a clinically significant lesion. The implemented registration algorithms took less than 30 seconds and 2 minutes for manually defined point- and normal vector features, respectively. The thesis concludes with a summary of potential applications and future research directions.
Supervisor: Barratt, D. ; Hawkes, D. Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available