Related Experiment Video
Updated: Jun 19, 2025
07:23
Retinal Vascular Reactivity as Assessed by Optical Coherence Tomography Angiography
Published on: March 26, 2020
7.5K
Anomaly Detection in Optical Coherence Tomography Angiography (OCTA) with a Vector-Quantized Variational Auto-Encoder
Hana Jebril1, Meltem Esengönül1, Hrvoje Bogunović1,2
1Lab for Ophthalmic Image Analysis, Department of Ophthalmology and Optometry, Medical University of Vienna, 1090 Vienna, Austria.
Bioengineering (Basel, Switzerland)
|July 27, 2024
Summary
This study introduces a novel deep learning model for detecting anomalies in Optical Coherence Tomography Angiography (OCTA) scans, aiding in the identification of retinal diseases and systemic conditions.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Optical coherence tomography angiography (OCTA) is crucial for assessing retinal blood flow.
- Abnormalities in retinal perfusion can signal ocular or systemic diseases.
Purpose of the Study:
- To develop and evaluate a deep learning-based anomaly detection model for OCTA images.
- To identify and localize anomalies indicative of potential health issues.
Main Methods:
- Utilized a combination of Vector-Quantized Variational Auto-Encoder (VQ-VAE) with Auto-Regressive (AR) modeling for representation learning.
- Employed Bayesian U-Net for epistemic uncertainty estimation in vasculature segmentation.
Main Results:
- Achieved an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.92 on the DRAC dataset and 0.75 on the OCTA-500 dataset for anomaly detection.
- Demonstrated effective anomaly localization with a mean Dice score of 0.61 on the DRAC dataset.
Conclusions:
- The proposed deep learning model effectively detects anomalies in OCTA scans.
- This represents the first work to address anomaly detection specifically within OCTA imaging.
