Fast upper airway MRI of speech

UnrestrictedMagnetic resonance imaging (MRI) is a powerful non-invasive imaging modality, but is relatively slow compared to alternatives such as X-ray and ultrasound. Accelerating MRI scans has been of the great interest over the past several years and the acceleration of upper airway MRI, in particular, is a primary focus of this thesis. Rapid and real-time MRI can be used to capture tissue dynamics (e.g., the tongue/velum during speech, or the beating heart) or to reduce scan time. Rapid MRI can be achieved through the use of novel acquisition and reconstruction methods. Acquisition technologies such as echo-planar imaging and spiral imaging are effective at improving temporal resolution, but introduce additional image artifacts. Image reconstruction techniques such as parallel imaging and compressed sensing accelerate acquisition speed by highly undersampling Fourier data and improve image quality by removing spatial aliasing artifacts.; In this dissertation, I present four methods that accelerate MRI and manage the associated artifacts. First, a new interleaved echo-planar imaging technique is presented and applied to real-time interactive cardiac MRI. When combined with parallel imaging reconstruction, this enables automatic correction of ghosting artifacts and maintains temporal resolution. Second, a method for accelerated three-dimensional (3D) upper airway imaging is presented. The method adopts compressed sensing for acceleration and enables extraction of a whole 3D vocal tract during a single trial of sustained sound production. Hence, it is free from image mis-registration problem and dramatically improves throughput in data acquisition. Third, a novel acquisition scheme based on a golden-ratio spiral temporal view order is presented and applied to real-time speech MRI. This provides more flexible retrospective selection of temporal resolution in vocal tract imaging than conventional spiral acquisition. I demonstrate its effectiveness at capturing rapid motion of articulators (e.g., tongue and velum) in nasal sound production. Finally, a novel 16-channel receive coil is described that is highly sensitive to the upper airway regions of interest (e.g., lips, tongue, soft palate). Its higher signal-to-noise ratio (SNR) and acceleration in parallel imaging is demonstrated by comparing it with widely available commercial coils