Principles of Optical Coherence Tomography AngiographyCHAPTER 1

Optical coherence tomography (OCT) has become part of the standard of care in ophthalmology. It provides cross-sectional and three-dimensional (3D) imaging of the anterior segment, retina, and optic nerve head with micrometer-scale depth resolution. Structural OCT enhances the clinician's ability to detect and monitor fluid exudation associated with vascular diseases. It, however, is unable to directly detect capillary dropout or pathologic vessel growth (neovascularization) that constitutes the major vascular change associated with two leading causes of blindness, age-related macular degeneration and diabetic retinopathy. These features, among other vascular abnormalities, are assessed clinically using fluorescein or indocyanine green (ICG) angiography. To overcome conventional structural OCT's inability to provide direct blood flow information, several OCT angiography methods have been developed.

Initially, Doppler OCT angiography methods were investigated for the visualization and measurement of blood flow.^1-6 Because Doppler OCT is sensitive only to motion parallel to the OCT probe beam, it is limited in its ability to image retinal and choroidal circulation, which are predominantly perpendicular to the OCT beam. An alternative approach has been speckle based OCT angiography. It has advantages over Doppler-based techniques because it uses the variation of the speckle pattern in time to detect both transverse and axial flow with similar sensitivities. Amplitude-based,^7-9 phase-based,¹⁰ or combined amplitude + phase¹¹ variance methods have been described.

We developed an amplitude-based method called split-spectrum amplitude-decorrelation angiography (SSADA). The SSADA algorithm detects motion in blood vessel lumen by measuring the variation in reflected OCT signal amplitude between consecutive cross-sectional scans. Decorrelation is a mathematical function that quantifies variation without being affected by the average signal strength, as long as the signal is strong enough to predominate over optical and electronic noise. This novelty of SSADA lies in how the OCT signal is processed to enhance flow detection and reject axial bulk motion noise. Specifically, the algorithm splits the OCT image into different spectral bands, thus increasing the number of usable image frames. Each new frame has a lower axial resolution that is less susceptible to axial eye motion caused by retrobulbar pulsation. This lower resolution also translates to a wider coherence gate over which reflected signal from a moving particle such as a blood cell can interfere with adjacent structures, thereby increasing speckle contrast. In addition, each spectral band contains a different speckle pattern and independent information on flow. When amplitude decorrelation images from multiple spectral bands are combined, the flow signal is increased. Compared to the full-spectrum amplitude method, SSADA using four-fold spectral splits improved the signal-to-noise ratio (SNR) by a factor of two, which is equivalent to reducing the scan time by a factor of four.¹² More recent SSADA implementations use even more than a four-fold split to further enhance the SNR of flow detection. As shown by an example from en face angiograms of the macular retinal circulation collected using a commercial 70 kHz 840 nm spectral OCT (Figs 1.1A to H), SSADA provides a clean and continuous microvascular network and less noise just inside the foveal avascular zone (FAZ).

The en face angiograms were generated by projecting the maximum decorrelation or flow value for each transverse position within the segmented depth range, representing the fastest flowing vessel lumen in the segmented tissue layers. In healthy eyes, the retinal angiogram shows a vascular network around the FAZ. The layers of the retina and choroid can be more finely separated to provide additional information to define diagnostic parameters of vascular defects. This will be discussed in Chapter 2.

To determine how the decorrelation or flow signal produced by the SSADA algorithm relates to flow velocity, phantom experiments were performed.¹³ The study showed that SSADA is sensitive to both axial and transverse flow, with a slightly higher sensitivity for the axial component. For clinical retinal imaging where the OCT beam is approximately perpendicular to the vasculature, the SSADA signal can be considered to be independent of the small variation in beam incidence angle for all practical purposes. In addition, it was found that decorrelation was linearly related to velocity over a limited range. A higher decorrelation value thus implies higher velocity flow. This range is dependent on the time scale of the SSADA measurement. With a 70 kHz spectral OCT system and 200+ A-scans per cross-sectional B-scan, SSADA should be sensitive to even the slowest flow at the capillary level, where flow speeds have been estimated at between 0.4 to 3 mm/s.^14,15 In larger vessels with higher velocities, the SSADA signal reaches a maximum value (saturates).

Compared to fluorescein or ICG angiography, the gold standards of retinal vascular imaging, OCT angiography has a number of advantages and differences. SSADA can be acquired in a few seconds and does not require intravenous injection, whereas fluorescein or ICG angiography requires multiple image frames taken over several minutes and can cause nausea, vomiting, and, albeit rarely, anaphylaxis.¹⁶ The fast and noninvasive nature of OCT angiography also means that follow-up scans can be conducted more frequently.

OCT angiography has several limitations. First, shadowgraphic flow projection artifact makes the interpretation of en face angiograms of deeper vascular beds more difficult. These artifacts are a result of fluctuating shadows cast by flowing blood in a superficial vascular layer that cause variation of the OCT signal in deeper, highly reflective layers. The flow projection artifact from the retinal circulation can be seen clearly on the bright retinal pigment epithelium (RPE). This artifact can be removed by software processing. The projection from the retinal circulation is relatively sparse and can be removed from deeper layers fairly effectively. However, the choriocapillaris is nearly confluent, and its projection and shadow effects are difficult to remove from deeper choroidal layers. A second limitation is the fading of OCT and flow signal in large vessels due to the interferometric fringe washout effect associated with very fast blood flow, especially the axial flow component.¹⁷ This means that central retinal vessels in the disc and large vessels in the deep choroid cannot be visualized using SSADA. Third, the scan area of OCT angiography is relatively small (3×3 to 6×6 mm). Larger-area angiograms of high quality can be achieved, but require higher speed OCT systems that are not yet commercially available.¹⁸ Lastly, because OCT angiography best resolves pathology when viewed as en face angiograms of anatomic layers, practical clinical applications require accurate segmentation software. Post-processing software is also needed to reduce motion and projection artifacts. The need for these sophisticated algorithms means OCT angiography still has much room to improve in the foreseeable future.

The SSADA algorithm was initially implemented on a custom-built 100 kHz 1050 nm wavelength swept-source OCT system. To generate high quality angiograms (Fig. 1.2A), 8 consecutive cross-sectional scans at each position were necessary. A scan pattern of 200 cross-sectional scan positions each with 200 axial scans was used. The overall angiographic scan pattern had 200 × 200 transverse points. 200 × 200 × 8 axial scans were acquired in 3.5 seconds.