Related Experiment Video
Updated: Feb 17, 2026
Author Spotlight: Understanding Age-Related Macular Degeneration Pathophysiology with QAF Workflow
Published on: May 26, 2023
Fully Automated Detection and Quantification of Macular Fluid in OCT Using Deep Learning
Thomas Schlegl1, Sebastian M Waldstein2, Hrvoje Bogunovic3
1Christian Doppler Laboratory for Ophthalmic Image Analysis, Department of Ophthalmology, Medical University Vienna, Vienna, Austria; Computational Imaging Research Lab Department of Biomedical Imaging and Image-Guided Therapy, Medical University Vienna, Vienna, Austria.
A new deep learning method accurately detects and quantifies macular fluid in OCT images for conditions like AMD and DME. This automated approach shows high accuracy, comparable to manual assessments, improving retinal diagnosis.
Area of Science:
- Ophthalmology
- Medical Imaging
- Artificial Intelligence
Background:
- Macular fluid detection is crucial for diagnosing exudative macular diseases.
- Current methods for fluid quantification can be time-consuming and subjective.
- Optical Coherence Tomography (OCT) is a key imaging modality for retinal assessment.
Purpose of the Study:
- To develop and validate a fully automated deep learning method for detecting and quantifying macular fluid in OCT images.
- To assess the accuracy and reliability of the automated method compared to manual expert grading.
- To evaluate the method's performance across different macular pathologies and OCT devices.
Main Methods:
- A deep learning algorithm was developed to automatically detect and quantify intraretinal cystoid fluid (IRC) and subretinal fluid (SRF).
- The dataset comprised 1200 OCT volumes from patients with neovascular age-related macular degeneration (AMD), diabetic macular edema (DME), and retinal vein occlusion (RVO).
- Algorithm performance was evaluated against manual consensus readings using metrics like Area Under the Receiver Operating Characteristics Curve (AUC), precision, and recall.
Main Results:
- The automated method achieved high accuracy for IRC detection (mean AUC 0.94) and SRF detection (mean AUC 0.92).
- Excellent correlation was found between automated and manual fluid quantification (Pearson's correlation coefficient of 0.90 for IRC, 0.96 for SRF).
- The method demonstrated superior performance in neovascular AMD and RVO compared to DME.
Conclusions:
- Deep learning-based automated analysis of OCT images provides accurate detection and quantification of retinal fluid.
- The automated method shows high concordance with manual expert assessment, enhancing diagnostic reliability.
- This technology offers a promising advancement for clinical practice and research in ophthalmology.
