(June 29) Seeing speech: In vivo mapping of anatomy and motion of the human tongue during speech using multimodal MRI

The human tongue is a volume preserving structure with highly complex, orthogonally oriented, inter-digitated muscles. This complex anatomy is necessary to move and deform the tongue throughout the vocal tract, which shapes the airway to produce swallowing and speech. However, this complex anatomy makes it extremely challenging to understand muscle specific roles and interactions among different muscles. Despite the growing need to accurately assess the similarities and differences of tongue muscular anatomy and motion patterns across different speakers, there has been a gap in our ability. Recent development in our lab of various MR technologies including high-resolution, cine, diffusion, and tagged MRI allows us to investigate tongue anatomy and internal tongue motion patterns. Multimodal image analysis, based on spatio-temporal imaging data, is becoming a popular approach to relate anatomy and function in the brain and the heart. Combining imaging modalities to study tongue structure and function is an appreciable challenge from the standpoint of inferring clinically and scientifically valuable information, because its deformations are so varied.

   In this presentation, I will outline our recent computational efforts toward developing state-of-the-art MR imaging and image analysis approaches to describe tongue anatomy and motion patterns during speech. Specifically, I will present methods to build a 3D vocal tract atlas and statistical models from structural MRI and a 4D atlas from cine-MRI to understand the standard anatomy and motion of the vocal tract during speech. When complete, this will be a comprehensive and systematic framework to characterize the relationship between tongue muscle structure and function during speech. Also, I will present a method to reveal functional organization of the tongue by determining functional units of the tongue motion. Taken together, the motion analysis framework has already been applied successfully to normal controls and glossectomy patients during speech.