Dynamic speech imaging with low-rank approximation

spatiotemporal modeling;partially separable functions;speech imaging;dynamic MRI;Magnetic resonance imaging (MRI)

Fu, Maojing ; Liang ; Zhi-Pei

关键词: spatiotemporal modeling;    partially separable functions;    speech imaging;    dynamic MRI;    Magnetic resonance imaging (MRI);   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/32073/Fu_Maojing.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Dynamic speech imaging is a powerful technique for real-time visualization of speech dynamics. As a promising modality for dynamic speech imaging, magnetic resonance imaging (MRI) can provide good soft-tissue contrast in an arbitrary imaging plane with a non-invasive procedure. However, conventional MRI suffers from low spatiotemporal resolution, which limits its applica-tion in dynamic speech imaging. This thesis presents a novel model-based dynamic MR imaging method to capture speech dynamics in high spatiotemporal resolution. Specifically, high spatiotemporal resolution reconstruction from very sparsely sampled data is achieved using the partial separability (PS) model, which takes advantage of the spatiotemporal correlations of dynamic speech images. The sampling pattern is also optimized to better capture speech dynamics. The spatial-spectral sparsity constraint is further incorporated into the basic PS model-based reconstruction to improve reconstruction quality. The effectiveness of the above approaches is demonstrated through systematic simulations and preliminary in vivo experiments.