Gastroesophageal Reflux Disease Diagnosis Using Hierarchical Heterogeneous Descriptor Fusion

• Main Authors: Yan-Ting Chen, 陳彥廷
• Other Authors: Chun-Rong Huang
• Format: Others
• Language: en_US
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/qchxhj

碩士 === 國立中興大學 === 資訊科學與工程學系所 === 101 === A new computer-aided diagnosis method is proposed to diagnose gastroesophageal reflux disease (GERD) from endoscopic images of the esophageal-gastric junction. To avoid the inferences of endoscope devices and automatic camera white balance adjustment, multiple color invariant models are used to represent endoscopic images. Then, visual vocabularies are built from each color model to describe the mucosa of the esophageal-gastric junction for support vector machine training. To simultaneously consider the prediction results of each color model, a hierarchical support vector machine scheme is proposed. During validation, visual vocabularies extracted from the test endoscopic image are used as the input of the hierarchical support vector machine to diagnose GERD. As shown in the experiments, our method can automatically diagnose GERD without any manual selection of region of interest and achieve better accuracy compared to methods using only one color invariant model.