Automated Beam Stitching and Segmentation Procedure for Space Division Multiplexing Optical Coherence Tomography Angiography
Year: 2025
Type: article
Abstract: Optical Coherence Tomography Angiography (OCTA) has revolutionized ophthalmic imaging and its capability to produce high-resolution 3D maps of the retinal microvasculature is instrumental in diagnosing retinovascular diseases such as diabetic retinopathy and age-related macular degeneration; However, the existing OCTA devices often suffer from slow acquisition speed limiting the field-of-view (FOV) in the clinic. Space Division Multiplexing OCTA (SDM-OCTA) address these limitations by acquiring multiple beams simultaneously, achieving manyfold faster acquisition speeds than single beam OCTA systems. But as each beam contains only part of the image, SDM-OCTA requires additional processing steps to produce coherent wide-field images. Though manual stitching and segmentation was previously used, it is time-consuming and prone to human error. The need for an automated solution is clear, especially in clinical settings where rapid and accurate imaging is essential for timely diagnosis and treatment of retinal diseases. Stitching and segmenting the multiple beams is a non-trivial task due to the misalignment that can occur during acquisition, and the large data sizes making it computationally expensive to process using traditional segmentation methods. This thesis presents an automated beam stitching and segmentation procedure for SDM-OCTA imaging, which offers benefits in the clinical setting by producing wide-field projections of retinal microvasculature quickly and accurately. This procedure utilizes a combination of custom graph-theoretic algorithms and deep learning techniques to integrate beam cropping, registration, stitching, flattening, and segmentation steps. Written in Rust, the implementation is designed to be fast and efficient, allowing for the processing of large SDM-OCTA volumes in a few seconds. We evaluate the performance of our method using public and private datasets, and demonstrate its utility in clinical settings.
DOI: 10.7936/sjq5-st12
Author: Song, Andrew
Language: English
Topic: Retinal Imaging and Analysis
Subfield: Radiology, Nuclear Medicine and Imaging
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7026820854
FWCI:
Cites:
Cited by:
Related to: