Related Experiment Video
Updated: Jul 16, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
Foveal avascular zone segmentation using deep learning-driven image-level optimization and fundus photographs.
I Coronado1, S Pachade1, H Dawoodally1
1Center for Precision Health, School of Biomedical Informatics, University of Texas Health Science Center at Houston (UTHealth), TX, USA.
Proceedings. IEEE International Symposium on Biomedical Imaging
|September 14, 2023
Summary
Estimating the foveal avascular zone (FAZ) from widely available fundus photos is feasible. This research demonstrates methods to approximate FAZ metrics, crucial for diagnosing retinal pathologies and assessing visual acuity.
Area of Science:
- Ophthalmology
- Medical Imaging
- Retinal Vascular Diseases
Background:
- The foveal avascular zone (FAZ) is critical for visual acuity and linked to retinal pathologies.
- Optical Coherence Tomography Angiography (OCT-A) accurately visualizes the FAZ but is limited to research settings.
- Fundus photography is widely accessible and utilized in population studies.
Purpose of the Study:
- To evaluate the feasibility of estimating the FAZ from standard fundus photographs.
- To develop and compare different computational approaches for FAZ estimation using fundus images.
Main Methods:
- Three methods were tested: pixel-level FAZ segmentation, image-level FAZ area regression, and a mask-free pipeline using saliency maps and active contours.
- The mask-free approach facilitates training without manual alignment of fundus and OCT-A images.
- Methods were validated against human-graded masks.
Main Results:
- Segmentation methods trained on pixel-level and image-level data showed good agreement with human annotations (DICE scores of 0.45 and 0.4, respectively).
- The study confirms the potential of fundus photography for FAZ estimation.
- A mask-free training pipeline was successfully implemented.
Conclusions:
- Fundus photography can serve as a viable proxy for estimating FAZ metrics when OCT-A is unavailable.
- These findings support the broader clinical application of fundus imaging for retinal health assessment.
- The developed methods offer a pathway for large-scale population studies on retinal vasculature.
