Related Experiment Video
Updated: Jul 15, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
Self-FI: Self-Supervised Learning for Disease Diagnosis in Fundus Images
Toan Duc Nguyen1, Duc-Tai Le2, Junghyun Bum3
1Department of AI Systems Engineering, Sungkyunkwan University, Suwon 16419, Republic of Korea.
Bioengineering (Basel, Switzerland)
|September 28, 2023
Summary
This study introduces a new self-supervised learning method for classifying ultra-wide-field fundus images (UFI), significantly improving disease detection. The approach uses contrastive learning to achieve state-of-the-art performance with less labeled data.
Area of Science:
- Medical Imaging
- Computer Vision
- Artificial Intelligence
Background:
- Self-supervised learning (SSL) shows promise in medical imaging.
- Labeled data for medical image classification is scarce and expensive.
- Ultra-wide-field fundus images (UFI) require effective classification methods.
Purpose of the Study:
- To develop a novel self-supervised learning method for UFI classification.
- To reduce the dependency on large labeled datasets in medical image analysis.
- To enhance disease detection accuracy in UFI.
Main Methods:
- Utilized contrastive learning for pre-training deep learning models.
- Employed bi-lateral contrastive learning fusing multiple image views.
- Incorporated multi-modality pre-training with conventional fundus images (CFI).
Main Results:
- Achieved state-of-the-art performance in UFI classification.
- Obtained an Area Under the ROC Curve (AUC) score of 86.96.
- Demonstrated superior performance compared to existing methods.
Conclusions:
- Self-supervised learning is a viable and effective approach for medical image analysis.
- The proposed method shows potential for improving clinical disease detection.
- SSL methods can overcome limitations of scarce labeled medical data.
