Skip to content Skip to navigation

PhD Defense: Machine Learning driven Large-Scale Medical Image Analysis and Its Applications in Cardiac Magnetic Resonance Imaging

Abstract: 

The analysis of left ventricle (LV) wall motion is an important step for understanding cardiac functioning mechanisms, and clinical diagnosis of ventricular diseases. For example, ventricular dyssynchrony is one of the major causes for heart failure; treatment of dyssynchrony, e.g. Cardiac Resynchronization Therapy (CRT), can help some patients preventing failure. Conventional diagnosis methods, including electrocardiogram (ECG) and ultrasound imaging, provide only coarse characterization of dyssynchrony patterns, such as global function indices or qualitative assessment of motion patterns. To achieve a more comprehensive understanding of ventricular dyssynchrony, we propose a novel approach to study the regional patterns of left ventricle (LV) wall using cardiac magnetic resonance imaging (MRI). Firstly, we extract the myocardial contours from long- and short-axis cine MRI, and compensate for respiration offsets through rigid transformation to reconstruct the 3D shell of the heart wall. Then an unsupervised learning method using deep neural networks is adopted to compute the in-plane deformation field. Next, the 3D volumetric LV wall motion and deformation fields are recovered by using deformable models and spatial interpolation. Finally, in order to characterize the regional motion of the LV wall, a conventional 17-segment model is utilized for dividing the reconstructed 3D model, so that the local dyssynchrony patterns can be well-determined. Our proposed approach has a great potential to be applied in the analysis of large-scale MRI datasets of various cardiovascular diseases, and used to guide the administration of CRT.

Speaker: 
Dong Yang
Location: 
CBIM 22
Event Date: 
04/05/2019 - 1:30pm
Committee: 
Prof. Dimitris Metaxas (Chair), Prof. Konstantinos Michmizos, Prof. Yongfeng Zhang, and Prof. James Duncan (Yale University)
Event Type: 
PhD Defense
Organization: 
Dept. of Computer Science