Related Experiment Video
Updated: Sep 25, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
Priors-guided convolutional neural network for 3D foveal avascular zone segmentation
Optics Express
|April 27, 2022
Summary
This study introduces a novel AI method for precise 3D segmentation of the foveal avascular zone (FAZ) in OCTA images. The approach improves accuracy in quantifying the FAZ, crucial for detecting retinal diseases.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- The foveal avascular zone (FAZ) is a key indicator of retinal health, particularly in the macula.
- Accurate 3D quantification of the FAZ is essential for diagnosing and monitoring various retinal pathologies.
Purpose of the Study:
- To develop an efficient and accurate 3D segmentation method for the FAZ in optical coherence tomography angiography (OCTA) images.
- To leverage a novel priors-guided convolutional neural network (CNN) for improved FAZ quantification.
Main Methods:
- Proposed a priors-guided CNN incorporating location and topology priors for 3D FAZ segmentation.
- Utilized a random central crop module and non-local attention gates for efficient processing and long-range dependency capture.
- Implemented a topological consistency constraint using persistent homology on projection maps to ensure prediction accuracy.
Main Results:
- The proposed method demonstrated significant reduction in over-segmentation compared to existing approaches.
- Achieved superior fitting to the contour of the FAZ region in OCTA images.
- Validated on two OCTA datasets comprising 478 eyes.
Conclusions:
- The developed CNN-based method offers a robust and accurate solution for 3D FAZ segmentation in OCTA imaging.
- This technique holds promise for enhancing the clinical assessment of retinal diseases through precise FAZ quantification.
