In this paper, we propose a method to extract sternocleidomastoid and longus capitis/colli automatically and measure the thickness of those muscles from cervical vertebrae ultrasound images. Extracting sternocleidomastoid is relatively easy but for longus capitis/colli case, due to the brightness sensitivity, it requires much more computationally burdensome procedures. In the binarization process, instead of simple and cheap thresholding method, we apply fuzzy sigma binarization to mitigate the sensitivity. Since that binarization procedure is computationally expensive we keep thresholding method for sternocleidomastoid case. With considerate image processing processes such as 4-directional contour analysis and Cubic Spline interpolation, we can successfully analyze features like muscle thickness.
In experiment, the efficacy of the proposed method is verified as having 73 ~ 87% of real world cervical vertebrae images successfully analyzed meaning that only a small magnitude of errors in measuring thickness from medical expert's own measurement (less than 0.1 cm for sternocleidomastoid and less than 0.3 cm for longus capitis/colli). We hope such result encourages the use of automatic ultrasound analysis system for cervical vertebrae in rehabilitation practices.
Keywords: Cervical vertebrae ultrasound images, colli muscle, longus capitis muscle, medical image processing, sternocleidomastoid, ultrasound images.