Related Experiment Video
Updated: Nov 20, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.1K
Efficient network selection for computer-aided cataract diagnosis under noisy environment
Turimerla Pratap1, Priyanka Kokil1
1Department of Electronics and Communication Engineering, Indian Institute of Information Technology Design and Manufacturing, Kancheepuram, Chennai 600127, India.
Computer Methods and Programs in Biomedicine
|January 23, 2021
Summary
A new computer-aided cataract diagnosis (CACD) method enhances robustness against image noise. This approach improves early cataract detection accuracy in noisy fundus retinal images.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Computer-aided cataract diagnosis (CACD) is vital for early detection.
- Existing CACD methods struggle with noise in fundus images, degrading performance.
- Noise is inherent in retinal image acquisition and transmission, necessitating robust methods.
Purpose of the Study:
- To propose an efficient, noise-robust CACD method for diagnosing cataracts.
- To address the performance degradation of current CACD techniques due to image noise.
Main Methods:
- A network selection-based robust CACD method using support vector networks (SVNs) trained at various noise levels.
- Automatic feature extraction from fundus images using a pre-trained convolutional neural network (CNN).
- Selection of an appropriate SVN based on the detected noise level in the input image.
Main Results:
- The proposed method demonstrated superior robustness against additive white Gaussian noise (AWGN) compared to existing CNN-based CACD methods.
- Analysis conducted on a dataset of good-quality and synthetically generated noisy fundus images.
Conclusions:
- The developed CACD method exhibits enhanced performance in noisy conditions.
- This research provides a foundation for future advancements in robust, CNN-based CACD techniques.
