INTERMEDIATE AND DEEP CAPILLARY PLEXUSES IN MACHINE LEARNING SEGMENTATION OF HIGH-RESOLUTION OPTICAL COHERENCE...

Area of Science:

• Ophthalmology
• Medical Imaging
• Biomedical Engineering

Background:

• Dye-based angiography has limitations in imaging the intermediate and deep capillary plexuses.
• High-resolution optical coherence tomography (HR-OCT) offers potential for detailed retinal vascular imaging.

Purpose of the Study:

• To describe imaging of the intermediate and deep capillary plexuses using machine learning-based segmentation of HR-OCT data.
• To evaluate the capability of HR-OCT to visualize retinal vascular layers not well-imaged by conventional angiography.

Main Methods:

• Spectral domain optical coherence tomography (SD-OCT) was used to image three healthy subjects.
• A random forest algorithm was employed for segmentation of the intermediate and deep capillary plexuses.
• Volume rendering was utilized for visualization of depth-resolved imaging data.

Main Results:

• HR-OCT successfully visualized the intermediate and deep capillary plexuses at the outer borders of the inner nuclear layer.
• These retinal vascular layers were depicted in unprecedented three-dimensional detail.
• Bridging vessels connecting to the deep capillary plexus's whorl-like patterns were observed, a feature previously seen only in histologic studies.

Conclusions:

• HR-OCT, combined with machine learning and advanced visualization, is highly relevant for studying retinal vasculature in health and disease.
• This noninvasive approach provides detailed, histology-like images of the deeper retinal vascular layers.
• The findings suggest a significant advancement in the noninvasive study of retinal microvasculature.