Title:Developing an Automatic Vision based Abdominal Muscle Extraction and Analysis Method from Ultrasonography
VOLUME: 12 ISSUE: 1
Author(s):Kwang B. Kim, Hyun J. Park, Doo H. Song and Byung-Kwan Choi
Affiliation:Department of Computer Engineering, Silla University, Busan 617-736, Korea.
Keywords:Muscle extraction, muscle analysis, abdominal ultrasonography, fascia features.
Abstract:Analyzing abdominal muscles and measuring useful associated morphological features is
important in many clinical researches. While ultrasonography is a non-invasive reliable tool for such
tasks, it may cause the experimenter dependent subjective diagnosis thus a computer vision based
automatic muscle detector/analyzer is much needed in this area. In this paper, we propose such an
automatic vision based method using a series of image processing algorithms. The novelty of our
method is to extract internal oblique muscle from abdominal ultrasonographic image which was excluded in previous
study due to their irregular features. Previously, we used Mask searching method to restore vague part of abdominal image
but the third layer of muscle (transverse abdominis) was not clearly extracted because of the image distortion in the
process. In order to analyze muscle morphometric features like the thickness, the second layer (internal oblique muscle)
should also be extracted correctly thus we develop a new muscle extraction process that includes extracting the second
layer by using unsharp masking. Extraction of transverse abdominis is also improved by developing a new data structure
Both-Map in candidate search process. In that process, we save the morphological features of initially extracted muscle
area and later such information is used to restore the fascia area with various image processing techniques to extract target
muscle accurately. The effectiveness of the proposed method is verified by the practical diagnostic field of rehabilitation
in the extraction accuracy and the magnitude of the muscle thickness measurement error that lies within 0.02cm in more
than 60% of cases used in the experiment.
