深層学習を用いた非アルコール性脂肪肝炎における肝線維化の超音波自動診断システム

DOI Web Site

書誌事項

タイトル別名
  • An Automated Ultrasound Diagnosis System for Liver Fibrosis in Non-Alcoholic Steatohepatitis Using Deep Learning

説明

<p>Liver fibrosis is important information for diagnosing the prognosis of fatty liver. Diagnosis of the degree of fibrosis by ultrasound is non-invasive and cost-effective. However, it is difficult to evaluate the effect of fat on interpretation and mild fibrosis. In this report, we propose a novel method utilizing deep learning to improve the accuracy and automation of ultrasound diagnosis of liver fibrosis for NASH. This is a novel system that extracts the parenchyma of liver by U-Net, and then performs classification using the network that considers the order of the fibrosis level. The experimental results showed that the extraction of parenchymal liver achieved a Dice coefficient of 0.929, demonstrating the effectiveness of the method using U-Net. As for the classification, the accuracy rate was improved to 0.639 than that of the conventional method.</p>

収録刊行物

キーワード

詳細情報 詳細情報について

問題の指摘

ページトップへ