Classification for Fatty Liver by Analysis Shear Wave Elastography Images

• Main Authors: HUANG, SHIH-TING, 黃士庭
• Other Authors: LEE, WEN-LI
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/pp674e

碩士 === 慈濟科技大學 === 放射醫學科學研究所 === 107 === B-mode ultrasound is most commonly used in classification for fatty liver in clinical examination. Some limitations are found when B-mode ultrasound is used. Operator dependence and subjective evaluation are factors which may cause errors in classification for fatty liver. This study used the functional ultrasound, SWE (shear wave elastography), for examination. We quantified SWE images and combined with serum fibrosis marker to establish four modules for classification fatty liver. In this study, clinical information, SWE images and serum fibrosis markers, of 28 subjects was collected. B-mode images were used as a grading criteria. The objects include 9 healthy subjects and 19 subjects with fatty liver. All the collected images were analyzed by image analysis software, ImageJ. The analytical parameter included R, G, B color ratio, no color ratio and the texture analysis of the SWE images. In addition, some measured values were obtained from clinical images. The aim of the study is to establish modules for classification of fatty liver by means of image analysis and statistical analysis. Four modules for classification of fatty liver were established: color ratio module, elastic texture module, combined analysis module and clinical measurement parameter module. The AUC were 0.867±0.024, 0.839±0.015, 0.911±0.106 and 1±0, respectively. All the classification modules’ AUC are over 0.8, which means all the classification modules are accurate and reliable. It is recommended to use the clinical measurement module for fatty liver classification. Two reasons are depicted. First, the AUC of the module are the highest. Second, the parameters used by this module are obtained directly from the clinic examination without any post processing on the images. Therefore, it is a convenient and accurate classification module.