Related Experiment Video
Updated: Oct 25, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.1K
Deep learning-enabled ultra-widefield retinal vessel segmentation with an automated quality-optimized angiographic
Duriye Damla Sevgi1, Sunil K Srivastava1, Charles Wykoff2
1The Tony and Leona Campane Center for Excellence in Image-Guided Surgery and Advanced Imaging Research, Cole Eye Institute, Cleveland Clinic, Cleveland, OH, USA.
Eye (London, England)
|August 10, 2021
Summary
A new deep learning tool can automatically segment retinal vasculature in ultrawidefield fundus angiography (UWFA) images. This technology aids in selecting optimal phase-specific images for consistent and objective evaluation of retinal vascular diseases.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Retinal vascular characteristics are dynamic and sensitive to imaging parameters.
- Accurate assessment of retinal vasculature is crucial for diagnosing and monitoring diseases like diabetic retinopathy and sickle cell retinopathy.
Purpose of the Study:
- To assess the feasibility of a deep learning (DL) based vascular segmentation tool for ultrawidefield fundus angiography (UWFA).
- To evaluate the DL tool's capability in automatically identifying quality-optimized, phase-specific images for retinal vascular analysis.
Main Methods:
- Deep learning was used for vascular segmentation of UWFA sequences.
- Cubic splines analyzed serial vascular changes across angiographic phases.
- The optimal early phase image was identified by maximum retinal vessel area (RVA), and a comparable late phase image was selected at ≥4 minutes.
Main Results:
- The DL tool was evaluated on 13,980 UWFA sequences (462 sessions).
- Optimal early and late phase images were successfully identified in 85.2% of sessions, with individual success rates of 90.7% (early) and 94.6% (late).
- Maximum RVA detection times varied among diabetic retinopathy, sickle cell retinopathy, and normal retinas.
Conclusions:
- Deep learning algorithms can extract vascular parameters for quality assessment and phase selection in UWFA.
- A DL-based system can enhance the speed, consistency, and objectivity of UWFA evaluations.
- This technology holds potential for improved clinical assessment of retinal vascular conditions.
