Data-Driven Biomedical Analysis, Modelling and Validation

Xing, Qi. George Mason University, ProQuest Dissertations Publishing, 2019

[link]

Abstract

Modern medical imaging techniques including Computed Tomography (CT), Magnetic Resonance Imaging (MRI), Ultrasound (US) and Microscope Imaging provide numerous images containing sufficient information. However, the clinical information is sealed under the pixels, textures and voxels and generally is not easy to be accessed. By integrating clinical observations and experiment data, biomedical image analysis and modeling approach can assist clinical application for disease diagnoses and treatment. The main objective of this dissertation is to develop quantitative tools that can assist clinicians in the diagnosis and treatment of patients. I investigate several computational approaches in addressing five biomedical problems. I make five main contributions through these biomedical problems: 1) Develop an automatic and objective method to segment extraocular muscles from magnetic
resonance images and measure extraocular muscles deformation from ultrasound images; 2) Design quantitative method to reconstruct and evaluate 3D human eyeballs; 3) Implement and validate novel method that combines different similarity measurements to optimally register volumetric ultrasound image data containing significant and local anatomical differences. I apply the method on 3D endovaginal ultrasound data acquired from patients during the biopsy procedure of the levator ani muscle; 4) Propose automatic heartbeat detection method to detect the heartbeat rate from body deformation in low resolution and low frequency video. The method is applied to track the heartbeats of immobilized, ventrally-positioned zebrafish larvae without direct larva heart observation; 5) Present a real time haptic spine surgical simulator that will be used to simulate spine surgeries with the advantages of being interactive, low-cost and representative.