摘要

Accurate measures of liver fat content are essential for investigating hepatic steatosis. For a non-invasive inexpensive ultrasonographic analysis, it is necessary to develop a fully automated reliable computer-aided software that can assist medical practitioners without any operator subjectivity. Among three typical stages of such software tool-extracting region of interest (ROI), extracting feature set, and classification for diagnosis-in this study, we attempt to develop fully automatic vision based liver and kidney area extraction subsystem. Actually, this extracting ROI stage usually requires manual procedure or human intervention with computer aided tools. Thus, our vision based fully automatic approach that extracts the liver area and the kidney area separately is rare to find. In experiment with human medical experts' verification, our software was successful to extract the liver and the kidney area with 87.5%similar to 95% in accuracy. Such findings will be useful in building reliable computeraided diagnostic software if combined with a good set of other characteristic feature sets and powerful machine learning classifiers in the future.

  • 出版日期2017-6