In a recent comprehensive review, researchers from the Casey Eye Institute at the Oregon Health and Science University outlined significant advancements—both hardware and software innovations—in optical coherence tomography angiography since 2020. These developments have improved scan speed, image quality, quantification capabilities, and diagnostic precision.

Imaging Performance

The adoption of Fourier domain processing and swept-source optical coherence tomography (SS-OCT) systems have enabled scan rates in the MHz range, so that commercial systems can reach 400 kHz and experimental systems can reach up to 9.4 MHz. These faster speeds allow larger single-shot fields of view (eg, 23 × 12 mm), improved sampling density, and reduced procedure time. The Fourier domain approach also allows flow signal in retinal microvasculature to be differentiated from confounding sources of motion, such as ocular pulsation, the investigators wrote in their review, which was published recently in Translational Vision Science & Technology.

Commercial platforms also now provide wide-field OCT angiography (OCTA; up to 26 × 21 mm), typically with some downsampling.

To counteract eye motion artifacts during long or wide-field scans, the investigators noted, researchers have developed self-tracking and passive rescanning approaches. Volumetric motion correction using orthogonal scans and real-time processing algorithms now improve image stability.

High transverse resolution has been achieved using adaptive optics or high-density sampling to enable visualization of capillaries as small as about 5 µm and reduce projection artifacts.

High-dynamic range OCTA extends flow detection sensitivity across a wider velocity range using multiple interscan times and variable interscan time analysis, which allows for the quantification of capillary flow speed.

Handheld SS-OCTA systems with wide-field imaging (up to 140°) and optimized ergonomics have been developed for use in infants and supine patients to support high-speed scanning and motion correction.

Image Processing and Quantitative Analysis

The investigators described advanced algorithms that enable volumetric projection artifact removal and statistical noise reduction. They explained that deep learning has further improved denoising without introducing artificial features.

Deep learning models can also now automatically detect and quantify several pathologies:

• Nonperfusion area, which is strongly correlated with diabetic retinopathy (DR) and glaucoma severity.

• Flow deficits in choriocapillaris, which predict development and progression of age-related macular degeneration (AMD) features like macular neovascularization (MNV) and geographic atrophy.

• Microaneurysms, which are identified by axial/transverse location and flow characteristics.

• Retinal neovascularization, which can be detected above the internal limiting membrane, often more effectively than dye-based angiography.

AI models trained on OCTA images have demonstrated high accuracy in:

• Diagnosing DR and ischemia

• Segmenting MNV lesions

• Detecting DR, AMD, and glaucoma from a single OCTA data set

• Differentiating arterial vs venous vasculature

• Inferring vascular leakage patterns comparable to fluorescein angiography.

Techniques like biomarker activation maps help visualize areas that influence artificial intelligence–based diagnosis for increased transparency in automated decisions, the investigators wrote.

Future Directions

The investigators noted that the development of larger data sets that have high-resolution images for deeper tissue layers and the largest fields of view, as well as images that measure flow speed, can achieve improved image analysis with accurate layer segmentation.

They concluded: “[I]mproved acquisition speeds [have] opened the way to new types of measurements (flow speed) and detection sensitivity and dynamic range. [F]iner resolution and larger fields of view mean we can measure vascular anatomy more extensively. Because of these trends we speculate that OCTA will be increasingly used as a clinical and research technology.”

Disclosures can be found in the published study.