Le, Than-Dung, Jouvet, Philippe et Noumeir, Rita.
2025.
« Improving transformer performance for french clinical notes classification using mixture of experts on a limited dataset ».
IEEE Journal of Translational Engineering in Health and Medicine, vol. 13.
pp. 261-274.
Prévisualisation |
PDF
Noumeir-R-2025-31059.pdf - Version publiée Licence d'utilisation : Creative Commons CC BY-NC-ND. Télécharger (5MB) | Prévisualisation |
Résumé
Transformer-based models have shown outstanding results in natural language processing but face challenges in applications like classifying small-scale clinical texts, especially with constrained computational resources. This study presents a customized Mixture of Expert (MoE) Transformer models for classifying small-scale French clinical texts at CHU Sainte-Justine Hospital. The MoE-Transformer addresses the dual challenges of effective training with limited data and low-resource computation suitable for in-house hospital use. Despite the success of biomedical pre-trained models such as CamemBERT-bio, DrBERT, and AliBERT, their high computational demands make them impractical for many clinical settings. Our MoE-Transformer model not only outperforms DistillBERT, CamemBERT, FlauBERT, and Transformer models on the same dataset but also achieves impressive results: an accuracy of 87%, precision of 87%, recall of 85%, and F1-score of 86%. While the MoE-Transformer does not surpass the performance of biomedical pre-trained BERT models, it can be trained at least 190 times faster, offering a viable alternative for settings with limited data and computational resources. Although the MoE-Transformer addresses challenges of generalization gaps and sharp minima, demonstrating some limitations for efficient and accurate clinical text classification, this model still represents a significant advancement in the field. It is particularly valuable for classifying small French clinical narratives within the privacy and constraints of hospital-based computational resources. Clinical and Translational Impact Statement—This study highlights the potential of customized MoE-Transformers in enhancing clinical text classification, particularly for small-scale datasets like French clinical narratives. The MoE-Transformer's ability to outperform several pre-trained BERT models marks a stride in applying NLP techniques to clinical data and integrating into a Clinical Decision Support System in a Pediatric Intensive Care Unit. The study underscores the importance of model selection and customization in achieving optimal performance for specific clinical applications, especially with limited data availability and within the constraints of hospital-based computational resources.
| Type de document: | Article publié dans une revue, révisé par les pairs |
|---|---|
| Professeur: | Professeur Noumeir, Rita |
| Affiliation: | Génie électrique |
| Date de dépôt: | 30 juin 2025 20:30 |
| Dernière modification: | 08 août 2025 21:49 |
| URI: | https://espace2.etsmtl.ca/id/eprint/31059 |
Actions (Authentification requise)
 |
Dernière vérification avant le dépôt |