Related Experiment Video
Updated: Dec 6, 2025
10:53
Image-guided, Laser-based Fabrication of Vascular-derived Microfluidic Networks
Published on: January 3, 2017
10.2K
Reconstruction of high-resolution 6×6-mm OCT angiograms using deep learning
Min Gao1, Yukun Guo1, Tristan T Hormel1
1Casey Eye Institute, Oregon Health & Science University, Portland, OR 97239, USA.
Biomedical Optics Express
|October 5, 2020
Summary
A new deep learning method, HARNet, enhances 6×6-mm optical coherence tomographic angiography (OCTA) scans. This improves image quality, reducing noise and artifacts for better clinical assessment.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Standard optical coherence tomographic angiography (OCTA) uses 3×3-mm or 6×6-mm acquisition areas.
- Larger 6×6-mm OCTA scans suffer from reduced quality, including lower signal-to-noise ratio and increased shadow artifacts due to undersampling.
Purpose of the Study:
- To introduce a deep-learning-based high-resolution angiogram reconstruction network (HARNet).
- To generate enhanced 6×6-mm superficial vascular complex (SVC) angiograms with improved image quality.
Main Methods:
- Trained HARNet on paired 3×3-mm and 6×6-mm OCTA datasets from the same eyes.
- Utilized deep learning for high-resolution angiogram reconstruction.
Main Results:
- Reconstructed 6×6-mm angiograms exhibited significantly lower noise intensity.
- Enhanced images showed improved contrast and vascular connectivity compared to original 6×6-mm scans.
- The algorithm avoided generating false flow signals at the original noise level.
Conclusions:
- HARNet effectively enhances the quality of 6×6-mm OCTA images.
- The image enhancement may lead to improved biomarker measurements and clinical assessments in ophthalmology.
- This deep learning approach addresses undersampling issues in larger OCTA field-of-views.
