Computer measurement of oral movement in swallowing

This article describes a computer program that automatically detects and tracks small metal markers affixed to a subject's tongue and teeth in fluoroscopic image sequences of swallowing. The program, written in Microsoft Visual C++ using Windows NT 4.0 and the SAVANT imaging toolkit, involves marker detection and marker tracking. Marker detection is done by template matching. A generic marker template was designed by extracting the grayvalues of pixels within an imaged marker. Template matching with a weighted center-of-mass calculation determined marker location with subpixel accuracy. Marker tracking employed a nearest-neighbor algorithm since (a) the movement of each marker was less than the distance between any two markers and (b) marker trajectories did not overlap. Effects of head motion were attenuated by computing tongue trajectories with respect to a constant frame of reference given by reference markers on the maxillary teeth. Motions were converted from pixels/frame to millimeters/second using a calibration ring secured to the subject's neck. Results for several image sequences showed that our program performs very well in terms of marker detection and trajectory determination. Comparison of automatic and manual tracking of the same image sequences indicated a high degree of correspondence. Automatic tracking of oral movement by computer is a useful tool in kinematic studies of swallowing.