Related Experiment Video
Updated: Aug 4, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
Performance analysis of pretrained convolutional neural network models for ophthalmological disease classification
1Department of Biostatistics, Faculty of Medicine, Izmir Katip Celebi University, Izmir, Turkey.
Arquivos Brasileiros De Oftalmologia
|April 5, 2023
Summary
This study shows that pretrained convolutional neural network models can effectively classify eight common eye diseases from fundus images. ResNet50 demonstrated superior performance in identifying various conditions, highlighting its potential for automated ophthalmological diagnosis.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Automated detection of eye diseases from fundus images is crucial for early diagnosis and treatment.
- Convolutional neural network (CNN) architectures have shown promise in medical image analysis.
Purpose of the Study:
- To evaluate the classification performance of pretrained CNN models (VGG16, Inceptionv3, ResNet50) for eight distinct ophthalmological diseases using fundus images.
- To identify the most effective CNN architecture for classifying specific eye conditions.
Main Methods:
- Utilized a public dataset of 10,000 fundus images from 5,000 patients, covering eight diseases.
- Implemented and compared VGG16, Inceptionv3, and ResNet50 models using an adaptive moment optimizer.
- Augmented training data and divided the dataset into 70% training, 10% validation, and 20% testing sets.
Main Results:
- ResNet50 achieved the highest accuracy (97.1%) and best overall performance for cataract classification (AUC=0.964, score=0.903).
- VGG16 achieved 96.2% accuracy but had lower sensitivity compared to ResNet50.
- Specific architectures showed varying strengths in classifying different diseases.
Conclusions:
- Pretrained CNN architectures are capable of accurately identifying ophthalmological diseases from fundus images.
- ResNet50 is recommended for classifying glaucoma, cataract, hypertension, and myopia; Inceptionv3 for age-related macular degeneration and others; and VGG16 for normal and diabetic retinopathy.
