Because OCT angiography does not require the use of exogenous dye, physicians may expect this new method to be an alternative to dye angiography. Although the visualization of retinal capillaries as seen in OCT angiography appears to be similar to that of FA, there are some essential differences in the resulting clinical information between the methods. In OCT angiography, dye leakage and pooling, which correspond to blood plasma leakage and pooling, cannot be detected. This is a critical disadvantage of OCT angiography, because dye leakage and pooling are clinically important findings to detect macular edema and active retinal and choroidal neovascularization (CNV) in the diagnosis and management of retinal and macular diseases. Thus, OCT angiography alone cannot be an alternative to dye angiography. However, serial OCT B-scans (a reflectivity intensity image) allow for the detection of intraretinal cystoid spaces and serous retinal detachment. Therefore, the use of angiographic OCT images taken together with OCT B-scan images enables us to manage retinal and macular diseases, even without FA or IA. OCT angiography allows for the depth resolution of vascular structures, which is not feasible in FA. This is expected to be a potentially powerful advantage of OCT angiography, as discussed in the next section. Importantly, because OCT angiography does not require the use of exogenous agents, it can be performed at every visit without the risk of allergic side effects. OCT angiography can provide additive clinical information that conventional angiography cannot; thus, it is recommended that OCT angiography be used in combination with conventional angiography for an accurate diagnosis, particularly for the diagnosis of rare diseases. However, the use of both reflectivity and angiographic OCT imaging may be enough to manage well-defined common pathologies, such as macular edema and serous retinal detachment, in eyes with diabetic maculopathy and retinal vein occlusive diseases.

The most powerful advantage of OCT angiography technology is the capability to resolve a 3D capillary network. The retinal capillary network is comprised of four layers: the retinal nerve fiber layer (RNFL), and the superficial, intermediate, and deep capillary plexus layers (Iwasaki and Inomata 1986; Snodderly and Weinhaus 1990; Snodderly et al. 1992). It is possible that the formation of retinal capillary pathologies in each layer, such as capillary non-perfusion and microaneurysms, differs among cases, and these differences may influence visual prognosis and response to treatment. In fact, retinal artery occlusive diseases reportedly showed different and variable involvement of the superficial and intermediate/deep capillary plexus layers (Yu et al. 2015). In particular, 22 % of eyes with retinal artery occlusive diseases had isolated deep capillary ischemia producing paracentral acute middle maculopathy while sparing the superficial capillary plexus, but the fluorescein angiographic appearance was normal (Yu et al. 2015). Although these findings were obtained from the increased reflectivity in each retinal layer as an indirect sign of ischemia, OCT angiography enables the direct observation of capillary perfusion and non-perfusion.

We have tried to resolve each of the four retinal capillary plexus layers based on a retinal layer segmentation method (Figs. 6.3 and 6.4). A raster scan for 3D OCT imaging can provide both reflectivity intensity and motion contrast signals that correspond to each other spatially. A segmentation algorithm to detect layer boundaries on reflectivity intensity B-scans allows for the determination of boundaries between the RNFL and ganglion cell layer (GCL), between the inner plexiform layer (IPL) and inner nuclear layer (INL), and between the INL and outer plexiform layer (OPL). Angiographic images between the posterior boundary of the RNFL and the line of x-pixels (1 pixel = 31.4 μm) anterior to the RNFL are defined as the RNFL capillary plexus (RNFLP) (Fig. 6.3). Angiographic images between the two lines of y-pixels anterior and posterior to the anterior boundary of the INL are defined as the intermediate capillary plexus (ICP). Angiographic images between the two lines of z-pixels anterior and posterior to the posterior boundary of the INL are defined as the deep capillary plexus (DCP). Angiographic images between the RNFLP and ICP are defined as the superficial capillary plexus (SCP). Capillary patterns on en face images were compared by changing the numbers of x, y, and z pixels.

Figure 6.4 presents en face images of the four retinal capillary plexus layers when x, y, and z were set to 5. The RNFL appearance shows a striated pattern of capillaries along the retinal nerve fiber bundles. The SCP, ICP, and DCP images each show different capillary network patterns. We found two major problems in depicting each of the four retinal capillary plexus layers. One was that the RNFLP around the fovea and temporal raphe could not be depicted. Figure 6.5 indicates that visualization of RNFLP appearance is improved when the value of x in Fig. 6.3 is changed from 1 to 6. At least 3 pixels are necessary to depict the RNFLP. However, even if the x value is increased to 6, the RNFLP around the temporal raphe cannot be visualized. This problem is likely attributable to the difficulty in segmenting the very thin RNFL around the fovea and temporal raphe (Fig. 6.6).

The second problem was a partial overlap of capillary patterns between the SCP and ICP (Fig. 6.7). This overlap is seen in and outside the fovea. This problem may be attributable to the anatomy of the retinal capillary plexus and the failure to detect motion-related signals in each layer selectively. The perifoveal capillary ring, as indicated by blue arrows in Fig. 6.7, is continuously connected with the SCP and ICP; that is, the perifoveal capillary network is not clearly shown to have four layers, and perifoveal capillaries are shared by the SCP and ICP. This anatomical overlap should be taken into consideration when the segmentation results are interpreted. The capillaries indicated by red arrows in Fig. 6.7 demonstrate that capillaries in the SCP are projected onto ICP images, regardless of how completely the SCP and ICP are separated spatially in the segmented volume. This problem may be a methodological failure in the OCT angiography algorithm. These overlaps could not be deleted by changing the value of y in Fig. 6.3 (Figs. 6.8 and 6.9). Appropriate values of y range from 1 to 5 for SCP imaging and from 5 to 11 for ICP imaging. We subtracted ICP signals at y = 1, 3, 5, 7, and 9 from those at y = 3, 5, 7, 9, and 11, respectively (Fig. 6.10). The subtracted images showed an SCP pattern when the subtraction was performed from ICP signals at y = 7 or more, suggesting that ICP images at y = 7 or more include SCP signals. Thus, the best value for y for both SCP and ICP imaging seems to be 5 for our algorithm. However, the current algorithm in our OCT angiography technique could not completely delete the partial signal overlap, even when the segmentation volume was changed. It appears that a z-value of at least 5 is required for clear DCP imaging (Fig. 6.11) that does not include an overlap with the ICP.

OCT angiography allows for the visualization of major pathological vascular features, including microaneurysms and capillary non-perfusion, in various retinal diseases such as diabetic retinopathy, retinal vein occlusion, and macular telangiectasia (MacTel) (Ishibazawa et al. 2015; Jia et al. 2015). A major interest is differences in the imaging of these features between FA and OCT angiography.