Related Experiment Video
Updated: Jun 9, 2025
09:17
Using Retinal Imaging to Study Dementia
Published on: November 6, 2017
21.6K
Clinical Utility of Deep Learning Assistance for Detecting Various Abnormal Findings in Color Retinal Fundus Images:
Joo Young Shin1, Jaemin Son2, Seo Taek Kong2
1Department of Ophthalmology, Seoul Metropolitan Government Seoul National University Boramae Medical Centre, Seoul, Republic of Korea.
Translational Vision Science & Technology
|October 23, 2024
Summary
Deep learning algorithms enhance ophthalmologists' ability to detect retinal abnormalities, improving sensitivity and reducing reading time, especially for less experienced readers. This technology aids in more accurate and efficient eye care diagnostics.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Retinal fundus photography is crucial for diagnosing various eye conditions.
- Inter-reader variability in interpreting these images can affect diagnostic accuracy.
- Deep learning (DL) shows promise in assisting medical image analysis.
Purpose of the Study:
- To assess the clinical utility of a DL-based detection tool for multiple retinal abnormalities.
- To evaluate the impact of DL assistance on readers with different levels of expertise.
Main Methods:
- Fourteen ophthalmologists (residents and specialists) reviewed 399 fundus images.
- Images were assessed with and without DL algorithm assistance in two separate sessions.
- Key metrics included sensitivity, specificity, and reading time per image.
Main Results:
- Algorithmic assistance significantly improved sensitivity for all 12 findings (P < 0.05).
- DL reduced inter-reader variance in sensitivity across all findings.
- Reading time decreased, particularly for residents and less complex images, with an estimated 25.9% reduction for residents.
Conclusions:
- Deep learning-based computer-assisted detection devices enhance diagnostic sensitivity and reduce reading time.
- These tools minimize inter-reader variability, leading to more consistent interpretations.
- DL assistance is particularly beneficial for less experienced clinicians in retinal image analysis.
