Sixth International Workshop on Models and Analysis of Vocal Emissions for Biomedical Applications (MAVEBA 2009)

Florence, Italy
December 14-16, 2009

Segmentation of Laryngeal High-Speed Videoendoscopy in Temporal Domain Using Paired Active Contours

Habib J. Moukalled (1,2), Dimitar D. Deliyski (1,2), Raphael R. Schwarz (1,3), Song Wang (2)

(1) Department of Communication Sciences and Disorders, University of South Carolina, Columbia, SC, USA
(2) Department of Computer Science and Engineering, University of South Carolina, Columbia, SC, USA
(3) Siemens AG, Erlangen, Germany

This paper introduces a method for segmention of the vocal-fold edges in temporal domain from laryngeal high-speed videoendoscopy (HSV). The method employs a pair of active contours (snakes), which deform within a series of kymographic images derived from the HSV data. By following a set of deformation rules, this pair of active contours converges to the desired boundaries of the glottis. The proposed method was tested on a dataset of 98 HSV samples, of which 96 were successfully segmented. The new method substantially outperforms existing methods. However, more precise analysis revealed that of the 96 successfully segmented HSV samples, 18 exhibited a fine error up to ±1 pixel, and 78 samples exhibited errors exceeding a pixel. The large majority of the gross errors (76%) were due to problems near the posterior and anterior commissures, which warrants further investigation for improving the accuracy and reliability of the method.

Index Terms. high-speed videoendoscopy, active contour segmentation, snakes, glottis, digital kymography

Full Paper (reprinted with permission from Firenze University Press)

Bibliographic reference.  Moukalled, Habib J. / Deliyski, Dimitar D. / Schwarz, Raphael R. / Wang, Song (2009): "Segmentation of laryngeal high-speed videoendoscopy in temporal domain using paired active contours", In MAVEBA-2009, 137-140.