MedWGAN Based Synthetic Dataset Generation for Uveitis Pathology - IMT Mines Albi-Carmaux Accéder directement au contenu
Article Dans Une Revue Intelligent Systems with Applications Année : 2023

MedWGAN Based Synthetic Dataset Generation for Uveitis Pathology

Résumé

Clinical decision support based on artificial intelligence (AI) methods have increasingly been employed in medical applications to support medical diagnosis. Developing efficient AI methods, however, depends necessarily on the availability of sufficiently large amount of data to provide reliable results. But, in medicine, it is not always possible to find sufficient amount of real data on all pathologies, particularly, for rare diseases. This paper proposes a methodological framework for generating synthetic data using data augmentation techniques combined with epidemiological profiles. It focuses on Uveitis, a rare disease in ophthalmology, which is difficult to diagnose because of the disparity in prevalence of its etiologies. The generated synthetic data have been qualitatively validated by specialist ophthalmologists and quantitatively tested using machine learning methods. Results show that, of a randomly selected sample of the generated data, more than 55% were assessed as good or excellent, which is very promising for generating synthetic, validated as near-real, medical data for rare diseases. They also show that the proposed framework is consistent in generating synthetic data, for Uveitis pathology, of different dataset sizes, achieving more than than 80% diagnosis prediction accuracy for 2000 patient records or larger.
Fichier principal
Vignette du fichier
MedWGAN-based-synthetic-dataset-generation-for-Uveitis-pathology.pdf (1.95 Mo) Télécharger le fichier
Origine : Fichiers éditeurs autorisés sur une archive ouverte

Dates et versions

hal-04067238 , version 1 (26-04-2023)

Licence

Paternité - Pas d'utilisation commerciale - Pas de modification

Identifiants

Citer

Heithem Sliman, Imen Megdiche, Loay Ajramy, Adel Taweel, Sami Yangui, et al.. MedWGAN Based Synthetic Dataset Generation for Uveitis Pathology. Intelligent Systems with Applications, 2023, 18, pp.200223. ⟨10.1016/j.iswa.2023.200223⟩. ⟨hal-04067238⟩
170 Consultations
65 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More