Noninvasive choroidal vessel analysis using deep learning: A novel approach to OCT angiography

A team of researchers from Peking University have developed an innovative, noninvasive choroidal angiography technique that enables layer-wise visualization and assessment of choroidal vessels using deep learning. (Image credit: Yanye Lu, Peking University)

Researchers from Peking University have developed an innovative, noninvasive choroidal angiography technique that enables layer-wise visualization and assessment of choroidal vessels using deep learning.

This method, published in Health Data Science, uses an advanced segmentation model capable of processing various qualities of optical coherence tomography (OCT) B-scans, making it a promising clinical tool for diagnosing retinal diseases.¹

Choroidal optical coherence tomography angiography (C-OCTA) represents a significant advancement in the study of choroidal vessels, which are key in conditions like age-related macular degeneration and central serous chorioretinopathy. Traditional methods, such as indocyanine green angiography (ICGA), are invasive and lack the volumetric data needed for detailed choroidal analysis.

The new method, proposed by Lei Zhu, a PhD student in the Department of Biomedical Engineering at Peking University, and Associate Professor Yanye Lu from the Institute of Medical Technology at Peking University Health Science Center, utilizes a deep learning framework to noninvasively capture three-dimensional choroidal vascular details from OCT B-scans.

“Our approach applies a segmentation model to extract choroidal vessels from OCT B-scans, trained on high-quality scans but also effective on lower-quality scans frequently seen in clinical practice. This enables more precise reconstruction of choroidal structures,” Zhu said. “It opens the door to improved analysis of choroidal indexes in various retinal diseases.”

The framework addresses the segmentation task as a cross-domain challenge, using an ensemble discriminative mean teacher structure to reduce noise and enhance adaptation between high-quality and low-quality B-scans. Testing on extensive datasets yielded a dice score of 77.28 for choroidal vessel segmentation, indicating its capability to accurately reconstruct choroidal vessel distributions.

In their study, the team demonstrated a notable reduction in vascular indexes in patients with central serous chorioretinopathy compared to healthy individuals, particularly in regions beyond the central macula fovea (P < 0.05). This work underscores the potential of this method for noninvasive clinical analysis and choroidal disease diagnosis.

The research team plans to expand the method to analyze choroidal indexes across a broader range of retinal diseases, potentially enhancing clinical precision and improving patient outcomes.

“Our goal is to continue refining this model for wider clinical use, offering a more efficient and less invasive alternative for choroidal vessel analysis,” added Lu.

Reference

1. Lei Zhu, Junmeng Li, Yicheng Hu, Ruilin Zhu, Shuang Zeng, Pei Rong, Yadi Zhang, Xiaopeng Gu, Yuwei Wang,Zhiyue Zhang, et al. Choroidal Optical Coherence Tomography Angiography: Noninvasive Choroidal Vessel Analysis via Deep Learning. Health Data Sci. 2024;4:0170.DOI:10.34133/hds.0170