Color fundus image enhancement - A deep learning based desktop app for earlier screening of diabetic retinopathy using real-time handheld fundus camera - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Color fundus image enhancement - A deep learning based desktop app for earlier screening of diabetic retinopathy using real-time handheld fundus camera

Zineb Sadok
  • Fonction : Auteur
Mohamed Akil
Anas Mansouri
  • Fonction : Auteur
Ali Ahaitouf
  • Fonction : Auteur
  • PersonId : 1110789

Résumé

Diabetic retinopathy (DR) is a common complication of diabetes that affects the blood vessels in the retina. It is a leading cause of vision loss worldwide. Early detection and intervention are crucial in preventing irreversible damage to the eyes. In this study, we propose a computer-aided diagnosis (CAD) system for the early detection of DR. Using a portable non-mydriatic fundus camera, we captured retinal images and applied deblurring and contrast enhancement techniques to improve image quality. We employed fine-tuning and transfer learning, specifically utilizing DenseNet-121, to detect DR from our private dataset. Additionally, two large datasets, APTOS and EyePACS, were used to train and evaluate different transfer learning DNN models. We found that the DenseNet-121 network achieves better results of accuracy with 97.3856 and 90.9000 respectively for APTOS and EyePACS datasets. The denseNet-121 is also used to detect DR from our private dataset and gives a higher accuracy of 98.6111. This work has designed a deep learning-based Desktop app, which captures and processes the fundus images for earlier screening of DR in remote medical centers or areas with limited access to Table-top fundus cameras and ophthalmologists.
Fichier principal
Vignette du fichier
EL-Mehdi_23.pdf (935.87 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04451313 , version 1 (11-02-2024)

Identifiants

  • HAL Id : hal-04451313 , version 1

Citer

El-Mehdi Chakour, Zineb Sadok, Mohamed Akil, Rostom Kachouri, Anas Mansouri, et al.. Color fundus image enhancement - A deep learning based desktop app for earlier screening of diabetic retinopathy using real-time handheld fundus camera. International Conference on Optimization and Data Science in Industrial Engineering (ODSIE 2023), Nov 2023, Istanbul, Turkey. ⟨hal-04451313⟩
17 Consultations
11 Téléchargements

Partager

Gmail Facebook X LinkedIn More