Development of AI-based fecal retention assessment method for rectal ultrasonography videos taken by home-visit nurses

DOI Open Access
  • Matsumoto Masaru
    Department of Imaging Nursing Science, Graduate School of Medicine, The Universityof Tokyo
  • Ishibashi Kousuke
    School of Integrated Health Sciences, Facultyof Medicine, The Universityof Tokyo
  • Kitamura Aya
    Department of Gerontological Nursing/Wound Care Management, Graduate School of Medicine, The Universityof Tokyo
  • Tamai Nao
    Department of Imaging Nursing Science, Graduate School of Medicine, The Universityof Tokyo Global Nursing Research Center, Graduate School of Medicine, The Universityof Tokyo
  • Miura Yuka
    Department of Imaging Nursing Science, Graduate School of Medicine, The Universityof Tokyo
  • Takahashi Toshiaki
    Department of Gerontological Nursing/Wound Care Management, Graduate School of Medicine, The Universityof Tokyo
  • Higashimura Shiho
    Department of Gerontological Nursing/Wound Care Management, Graduate School of Medicine, The Universityof Tokyo
  • Nakagami Gojiro
    Department of Gerontological Nursing/Wound Care Management, Graduate School of Medicine, The Universityof Tokyo Global Nursing Research Center, Graduate School of Medicine, The Universityof Tokyo
  • Sanada Hiromi
    Department of Gerontological Nursing/Wound Care Management, Graduate School of Medicine, The Universityof Tokyo Global Nursing Research Center, Graduate School of Medicine, The Universityof Tokyo

Bibliographic Information

Other Title
  • 訪問看護師が撮影した直腸エコー動画に対するAI による便貯留評価手法の考案

Search this article

Description

The purpose of this study was to develop a method for more accurate evaluation of fecal retention by an artificial intelligence interpretation support application that determines and displays in red extraction area the presence of fecal retention for rectal ultrasonographic videos taken by home-visiting nurses. In this study, static images were output from rectal ultrasonographic videos taken from medical records of one home-care nursing station. For each static image, the ultrasound expert and the application judged the presence of fecal retention based on half-moon or crescent-shaped hyperechoic findings as extraction area. This study focused on images that were false-positive results for the application when the ultrasound expert’s answer was correct. Extracted regions with a depth of less than 50 mm and a longitudinal diameter of less than 14.5 mm were excluded to distinguish them from stool and gas retentions outside the rectum. As a result, the sensitivity was 81.1%, specificity 87.8%, and correct rate 83.7%. A method of adjusting the depth and long diameter of the extraction area was developed as a means for improving the judgement accuracy of the application.

Journal

Related Projects

See more

Keywords

Details 詳細情報について

Report a problem

Back to top