Abstract:

Magnetic resonance imaging has served as a valuable tool for studying primarily static postures in speech production. Now, recent improvements in imaging techniques, particularly in temporal resolution, are making it possible to examine the dynamics of vocal tract shaping during speech. Examples include Mady et al. (2001, 2002) (8 images/second, T1 fast gradient echo) and Demolin et al. (2000) (4--5 images/second, ultra fast turbo spin echo sequence). The present study uses a non 2D-FFT acquisition strategy (spiral k-space trajectory) on a GE Signa 1.5T CV/i scanner with a low-flip angle spiral gradient echo originally developed for cardiac imaging [Kerr et al. (1997), Nayak et al. (2001)] with reconstruction rates of 8--10 images/second. The experimental stimuli included English sentences varying the syllable position of /n, r, l/ (spoken by 2 subjects) and Tamil sentences varying among five liquids (spoken by one subject). The imaging parameters were the following: 15 deg flip angle, 20-interleaves, 6.7 ms TR, 1.88 mm resolution over a 20 cm FOV, 5 mm slice thickness, and 2.4 ms spiral readouts. Data show clear real-time movements of the lips, tongue and velum. Sample movies and data analysis strategies will be presented. Segmental durations, positions, and inter-articulator timing can all be quantitatively evaluated. [Work supported by NIH.]