The focus over the past year has been in the development of quantitative feature extraction techniques based on our real-time submandibular ultrasound recordings. Previously, software has been developed to extract the tongue surface from ultrasound. Focus has now been placed on extraction of the hyoid bone region from ultrasound, as its trajectory is of primary clinical significance in the diagnosis of dysphagia. An image processing technique was developed to quantitatively extract the position of the hyoid from ultrasound images. The method utilizes the hyperechoic property of the many tissue inter-connections on and around the hyoid bone and makes use of a combination of spatial filters, gray scale enhancements, and image thresholding. An experiment was conducted utilizing simultaneously recorded ultrasound and videofluorographic images of swallowing to validate the hyoid extraction technique. Videofluorography is currently the """"""""gold standard"""""""" of swallowing studies. Currently, the hyoid extraction technique is being applied to ultrasound swallowing sequences stored in our archive. The major problem which must be overcome is the poor resolution of the scans produced by an antiquated ultrasound unit. Four swallowing studies have been digitized and analyzed using this technique. Further effort is being directed in the application of image processing techniques to recover information in the older scans. Recent advances in commercial ultrasound scanner technology have resulted in superior resolution of small parts such as the hyoid region as well as improved near field performance. We intend to use updated equipment in the next year and this will serve to improve the quantitative feature extraction techniques developed for speech and swallowing analysis.
