Pseudo-real-time retinal layer segmentation for high-resolution adaptive optics optical coherence tomography.

Area of Science:

• Ophthalmology
• Biomedical Imaging
• Medical Technology

Background:

• High-resolution imaging of retinal layers is crucial for diagnosing and monitoring eye diseases.
• Current segmentation methods for Sensorless Adaptive Optics-Optical Coherence Tomography (SAO-OCT) may not achieve real-time processing speeds.
• Real-time segmentation can enhance adaptive optics control for improved image quality.

Purpose of the Study:

• To develop and validate a pseudo-real-time retinal layer segmentation algorithm for high-resolution SAO-OCT.
• To enable the extraction of en face images during data acquisition for guiding optical coherence tomography system adjustments.
• To improve focus control and aberration correction in SAO-OCT systems.

Main Methods:

• A pseudo-real-time segmentation method based on Dijkstra's algorithm was implemented.
• The algorithm utilizes pixel intensity and vertical image gradients within a defined search region.
• Retinal layer boundaries are segmented iteratively based on prominence, segmenting up to six layers.

Main Results:

• The segmentation processing time is correlated with the number of retinal layers segmented.
• Average processing times for segmenting six layers in retinal B-scans (496×400 and 240×400 pixels) were 25.60 ms and 13.76 ms, respectively.
• Segmenting only two layers in a 240×400 pixel image required only 8.26 ms.

Conclusions:

• The developed pseudo-real-time segmentation method is efficient for high-resolution SAO-OCT.
• This technique facilitates real-time en face image extraction for adaptive optics control.
• The algorithm's speed and accuracy contribute to enhanced SAO-OCT system performance and potential clinical applications.