Development of Graph-Based Algorithm for Differentiating Pathophysiological Conditions

Authors

  • Satoshi IWAI The University of Tokyo
  • Tomohiro MITANI Department of Biomedical Informatics, Graduate School of Medicine, The University of Tokyo
  • Jin HAYAKAWA Department of Biomedical Informatics, Graduate School of Medicine, The University of Tokyo
  • Emiko SHINOHARA Department of Artificial Intelligence in Healthcare, Graduate School of Medicine, The University of Tokyo
  • Takeshi IMAI Center for Disease Biology and Integrative Medicine, Graduate School of Medicine, The University of Tokyo
  • Yoshimasa KAWAZOE Department of Artificial Intelligence in Healthcare, Graduate School of Medicine, The University of Tokyo
  • Kazuhiko OHE Department of Biomedical Informatics, Graduate School of Medicine, The University of Tokyo

Keywords:

Algorithms, Decision Support Techniques, Differential Diagnosis, Knowledge Bases System

Abstract

Aim: Clinical diagnostic decision support systems, which use pathophysiological information to improve diagnostic accuracy, have historically required knowledge of various relations between pathophysiological states to handle complex cases. Developing a knowledge model centered on pathophysiological functions instead of pathophysiological states may reduce this unwieldiness. Materials and Methods: In this study, such a knowledge model is provided by a modified and generalized factor graph, the pathophysiological query (PPQ) graph. A PPQ algorithm that automatically suggests possible pathological conditions of patients in the form of PPQ graphs is also developed. To evaluate the model and the algorithm, a computer software that processes the PPQ algorithm and PPQ graph, which represent the acid-base regulatory functions, was developed. Four case reports were considered, and up to two-time points, used as evaluation data points, were selected for each case. The software was used to obtain the diagnoses suggested by the PPQ model, which were then compared to diagnoses formulated by three physicians. Results: The output acquired by the proposed method was in accordance with the diagnosis of the physicians in three out of the five cases. Conclusion: The PPQ model may be a valuable diagnostic tool for suggesting differential pathological conditions to physicians in complex cases.

Downloads

Published

23.06.2020

How to Cite

1.
IWAI S, MITANI T, HAYAKAWA J, SHINOHARA E, IMAI T, KAWAZOE Y, OHE K. Development of Graph-Based Algorithm for Differentiating Pathophysiological Conditions. Appl Med Inform [Internet]. 2020 Jun. 23 [cited 2024 Jul. 21];42(2):107-1. Available from: https://ami.info.umfcluj.ro/index.php/AMI/article/view/774

Issue

Section

Articles