Related Experiment Video
Updated: May 31, 2026
09:17
Using Retinal Imaging to Study Dementia
Published on: November 6, 2017
Automated multi-level pathology identification techniques for abnormal retinal images using artificial neural
J Anitha1, C Kezi Selva Vijila, A Immanuel Selvakumar
1Department of ECE, Karunya University, Coimbatore 641114, India. rajivee1@rediffmail.com
The British Journal of Ophthalmology
|June 24, 2011
Summary
This study introduces an artificial neural network for accurate retinal image classification, aiding ophthalmologists. The efficient, time-saving tool achieved 97.7% accuracy in identifying four retinal disease categories.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Accurate classification of retinal diseases is crucial for timely treatment.
- Existing methods may lack efficiency or accuracy in multi-class identification.
Purpose of the Study:
- To develop an automated system using artificial neural networks for classifying abnormal retinal images.
- To achieve high accuracy and efficiency in identifying four specific retinal pathologies.
Main Methods:
- Utilized 420 abnormal retinal images across four categories: diabetic retinopathy, central retinal vein occlusion, central serous retinopathy, and neovascularization.
- Applied image pre-processing techniques including green channel extraction, histogram equalization, and median filtering.
- Employed texture-based feature extraction and Kohonen neural networks for pathology identification.
Main Results:
- Achieved an average classification accuracy of 97.7% (±0.8% deviation).
- Reported average sensitivity of 96% and specificity of 98%.
- The Kohonen neural network processed 420 images in an average of 300±40 seconds.
Conclusions:
- The developed approach serves as a valuable diagnostic tool for retinal disease identification.
- Artificial neural networks enable accurate simultaneous multi-level classification of abnormal retinal images.
- The time-efficient nature of the method is highly suitable for clinical ophthalmology applications.