Risk factor identification from clinical records for diabetic patients. - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2014

Risk factor identification from clinical records for diabetic patients.

Résumé

This paper presents the LIMSI experiments to identify risk factors in patient records, as part of our participation in the 2014 i2b2/UTHealth shared-task. Our approach relies on a CRF system complemented by hand-designed patterns to identify risk factor mentions, and on Weka's OneRule algorithm to compute the temporal attributes of each risk factor after transformation into a single-label classification task. We also used a tailored version of the Heideltime tool to identify temporal expressions and to normalize them according to the document creation time (DCT). Our best submission achieved a global .8451 micro F-measure. We obtained more accurate time attribute predictions for risk factors occurring after the DCT, and for medication before the DCT.
Fichier non déposé

Dates et versions

hal-02951136 , version 1 (28-09-2020)

Identifiants

  • HAL Id : hal-02951136 , version 1

Citer

Cyril Grouin, Véronique Moriceau, Sophie Rosset, Pierre Zweigenbaum. Risk factor identification from clinical records for diabetic patients.. 7th i2b2 Shared-Task and Workshop on Challenges in Natural Language Processing for Clinical Data, AMIA 2014 Affiliate Meeting: i2b2/UTHealth, Nov 2014, Washington DC, United States. ⟨hal-02951136⟩
30 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More