Visualizing electronic medical records of diabetic patients using pairwise similarity for explainable structuring - Archive ouverte HAL Access content directly
Conference Papers Year : 2020

Visualizing electronic medical records of diabetic patients using pairwise similarity for explainable structuring

Abstract

As medical databases grow larger and larger, medical experts oftenlack appropriate and accessible tools to make the best of the datasetsavailable and transform data into actionable information. Manyknowledge extraction algorithms provide relevant results but failto provide explainable and transparent results. Accountability isparamount in healthcare, and hospital staff cannot rely on black boxtools when it comes to taking informed decisions. To address thissituation we propose an algorithm able to structure thousands ofelectronic medical records by similarity and typicality. Using a rank-based approach suitable for high-dimensional data, we associateeach patient's record to a very similar yet more typical record.This provides a structure suitable for data visualization, allowingfor both a high-level summary of a cohort and its representativepatients, and a detailed representation of similarities and relationsin each cluster. We applied this method to electronic medical recordsof diabetic patients, providing an easy tool for visualization andexploration of these data with the added benefit of explainability.
Fichier principal
Vignette du fichier
SHIC_2020_paper_18.pdf (283.09 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-04085352 , version 1 (28-04-2023)

Identifiers

  • HAL Id : hal-04085352 , version 1

Cite

Joris Falip, Sara Barraud, Frédéric Blanchard. Visualizing electronic medical records of diabetic patients using pairwise similarity for explainable structuring. SHeIC 2020 - Smart Health International Conference 2020, May 2020, Troyes (France), France. ⟨hal-04085352⟩
14 View
8 Download

Share

Gmail Facebook X LinkedIn More