MS-LSTMEA: Predicting clinical events forn Hypertension using Multi-Sources LSTM Explainable Approach - Interactions, Réalité Augmentée, Robotique Ambiante Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2022

MS-LSTMEA: Predicting clinical events forn Hypertension using Multi-Sources LSTM Explainable Approach

Résumé

Hypertension is a major risk factor for cardiovascular disorders and diseases. Since it is expected to increase dramatically, effective Hypertension management becomes more and more critical. Early detection of patients with uncontrolled Hypertension would allow to employ personalized medicine for anticipating and providing the adequate drug class. In this paper, we propose MS-LSTMEA, a Multi-Source and Explainable Hypertension prediction Approach based on Long Short-Term Memory algorithm, for predicting both the drug class for patients and the date of their next medical appointment. MS-LSTMEA can successfully combine different sources of medical information about patients, represented by tabular data, while processing them separately and differently, and taking into account the temporal aspect of Electronic Health Records. In addition, it integrates an Attention mechanism which allows to improve the model results and to explain its outcomes. Experiments have been conducted on real Electronic Health Records data from 429,087 patients with the support of experts in healthcare. The results show significant improvements in accuracy over several existing models.
Fichier principal
Vignette du fichier
RR20221012_FZ_NS_NAM_Expert_System_With_Application_EHRs_Multi_sources.pdf (2.19 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03812885 , version 1 (13-10-2022)

Identifiants

  • HAL Id : hal-03812885 , version 1

Citer

Farida Zehraoui, Naziha Sendi, Nadia Abchiche-Mimouni. MS-LSTMEA: Predicting clinical events forn Hypertension using Multi-Sources LSTM Explainable Approach. 2022. ⟨hal-03812885⟩
36 Consultations
15 Téléchargements

Partager

Gmail Facebook X LinkedIn More