Optical coherence tomography angiography (OCTA) is a novel imaging method that allows non-invasive imaging of the retinal vasculature with resolution approaching that of histology.1–4 OCTA images retinal vessels by detecting variations in the intensity and/or phase properties of the OCT signal over multiple B-scans resulting from the movement of red blood cells.5 OCTA offers several advantages over fluorescein angiography in that it is faster, non-invasive, and has no side effects. Two studies have briefly surveyed diabetic retinopathy (DR) using spectral-domain OCTA (SD-OCTA). Schwartz et al. used a phase variance contrast method in one subject and found that impaired vascular perfusion was clearly shown whereas microaneurysms were variably seen.6 Jia et al. studied two subjects with DR using the intensity-based split-spectrum amplitude decorrelation angiography (SSADA) technique and applied a quantitative analysis to measure changes in vascular density.7 More recently, Ishibazawa et al. studied DR in 25 subjects using the SSADA approach with SD-OCT and described capillary drop-out, microaneurysms, and disc neovascularization.8
Although these recent studies of OCTA in DR have demonstrated a few of the vascular findings such as impaired capillary perfusion, microaneurysms, and neovascularization using SD-OCTA, there are important additional vascular findings in DR that are commonly observed in clinical practice that have not been described and may have diagnostic value if detectable by OCTA. In addition, the extent to which microaneurysmal findings are apparent in OCTA is not clear as compared to fluorescein angiography. For example, although Schwartz et al.6 and Ishibazawa et al.8 report that microaneurysms seen on OCTA do not always correlate with those on fluorescein angiography and vice versa, the degree of this discrepancy is not well-described, and the reasons for these discrepancies are not clear.
This study uses an intensity- and phase-based contrasting algorithm with a swept-source (SS)-OCTA device to demonstrate many of the clinically relevant findings in DR already reported, as well as findings that have not been reported, including intraretinal microvascular abnormalities (IRMA), cotton-wool spots, and some forms of intraretinal fluid. This study demonstrates that SS-OCTA is feasible and highly reliable for visualizing a broad spectrum of microvascular changes in DR and has the potential to be a clinically meaningful tool for assessing this disease.
Patients and Methods
The study was approved by the institutional review board of the University of Southern California, and a signed informed consent was obtained from each subject prior to enrollment. All subjects underwent standard clinical examination and testing as appropriate for their clinical disease. Fluorescein angiography was performed in subjects when clinically indicated using either a Spectralis (Heidelberg Engineering, Heidelberg, Germany) or Optos 200Tx (Optos, Dunfermline, Scotland) device. Thirty-three patients diagnosed with DR were recruited. Exclusion criteria was the presence of significant ophthalmological comorbidities such as retinal vein occlusion or severe media opacities. Ancillary imaging procedures were performed on the same day or within 1 month of the OCT angiograms used unless otherwise indicated.
OCT angiograms were acquired using a previously described1 Carl Zeiss Meditec (Dublin, CA) Cirrus prototype modified with a SS laser system with a central wavelength of 1,060 nm and a scan speed of 100,000 A-scans per second. Scans were acquired in a 3 mm × 3 mm pattern. B-scans were repeated four times over each section for assessment of intensity and phase differences. Each scan lasted approximately 4.5 seconds. Retinal vasculature was assessed within three horizontal retinal slabs of the OCTA consisting of the inner retina (internal limiting membrane, retinal nerve fiber layer, ganglion cell layer, and superficial inner plexiform layer), mid retina (deep inner plexiform layer, inner nuclear layer, outer plexiform layer, and superficial outer nuclear layer), and outer retina (deep outer nuclear layer to the external limiting membrane). The vasculature within each retinal slab was reconstructed using a phase- and intensity-based contrast algorithm and visualized as separate en face images.
Comparisons of OCT angiograms and available fluorescein angiograms were made using direct overlays created with Photoshop Elements 9.0 (Adobe Systems, San Jose, CA). In most cases, comparisons were made with fluorescein angiograms taken within the first 30 seconds of the transit. In some cases where this was not possible, later frames were used.
Impaired Vascular Perfusion
Areas of absent OCTA signal, signifying impaired vascular perfusion, were clearly visible in all cases where areas of capillary non-perfusion were also seen on fluorescein angiography. In many cases the extent of the impaired perfusion was more evident on OCTA than on fluorescein angiography at any point in the fluorescein transit (Figure 1F vs. 1B). OCTA allowed for clear delineation of the extent of impaired perfusion in the inner- and middle-retinal layers. Areas of impaired capillary perfusion in one layer were often associated with impaired perfusion in the other layer. However, these patches of impaired perfusion were rarely exactly the same between layers and some variation in shape and size were common (Figure 1F).
Fundus images and swept-source optical coherence tomography angiography (SS-OCTA) of a representative diabetic subject. Conventional color fundus (A) and zoomed-in fluorescein angiography of the macula (B) show areas of impaired perfusion, vascular looping and microaneurysms. (C) SS-OCT B-scan through dotted red line in (D) showing segmentation of the retina into the inner retina (blue to cyan line), mid-retina (cyan to red line), and outer avascular retina (red to yellow line). SS-OCTA of the same region as Figure 1B of the inner retina (D) and mid-retina (E) showing areas of impaired vascular perfusion (green arrowheads), intraretinal vascular looping (red arrowhead), an intraretinal microvascular abnormality (yellow arrow), and an irregular foveal avascular zone. (F) Depth-encoded SS-OCTA where red, green, and blue correspond to the inner-, mid-, and outer-retinal layers, respectively. (G) En face intensity SS-OCT scan of the full retina corresponding to the SS-OCTA in (D–F) does not demonstrate any artifactual decrease in OCT reflectivity.
OCTA clearly demonstrated impaired perfusion within cotton-wool spots (Figure 2E). The area of the cotton-wool spot was clearly visible as swelling of the nerve fiber layer on SD-OCT (Figure 2H), and it was evident that the majority of the inner capillary plexus was compromised in the area of the cotton-wool spot on depth-encoded pseudocolored images. Comparison of the area of impaired perfusion on OCTA and the size of the retinal nerve fiber layer swelling on SD-OCT shows excellent correlation between these findings.
Swept-source optical coherence tomography angiography (SS-OCTA) scans of diabetic subjects exemplifying typical findings seen in diabetic retinopathy. (A) Inner-retinal SS-OCTA of a diabetic subject with impaired vascular perfusion, intraretinal vascular looping (red arrowheads), and a microaneurysm (cyan arrowhead). (B) En face intensity SS-OCT scan of the inner retina corresponding to the SS-OCTA in above panel demonstrates some artifactual changes in reflectivity; however, the SS-OCTA findings are largely unchanged. (C) SS-OCT B-scan through red dotted line in (A) with segmentation of the inner retinal layer between the blue and cyan lines. (D) Fluorescein angiogram of the same subject and area seen in (A) with arrowheads from (A) overlaid on the image. (E) A cotton-wool spot is shown as an area of impaired vascular perfusion on a depth-encoded SS-OCTA and as a bulge in the nerve fiber layer on spectral-domain OCT (G). (F) En face intensity SS-OCT scan of the inner retina corresponding to the SS-OCTA in above panel demonstrates some artifactual changes in reflectivity; however, the SS-OCTA findings are largely unchanged. (H) SS-OCT B-scan through white dotted line in (E). (I) Inner-retinal section of a subject with neovascularization of the disc showing the neovascularization projecting out of the center of the disc. (J) Corresponding late phase fluorescein angiogram of the neovascularization.
Vascular Looping and Intraretinal Microaneurysmal Abnormalities (IRMA)
OCTA clearly showed looping vessels in many instances adjacent to areas of impaired capillary perfusion consistent with clinically defined IRMA (Figures 1D–F). Typically, the caliber of the loops was noticeably greater than that of surrounding capillaries. Vascular looping was not observed or easily evident in areas without impaired capillary perfusion. Although most vascular loops appeared of capillary origin (Figures 1D–F, 2A), in one case, a large-caliber vessel with looping extended into the preretinal space (Figure 3). The preretinal extension of the large-caliber vessel was contiguous with a single large-caliber intraretinal vessel on either end. In all cases that we observed, vascular looping extended across more than one retinal layer. These vessels generally did not leak in the late phase of fluorescein angiography and were not consistent with clinical appearance of neovascularization.
Large-caliber preretinal vascular looping seen in subject with severe diabetic retinopathy. (A) Red-free image with preretinal vascular loop in red gradient line. Magnified view of the preretinal loop is seen on infrared imaging (B) and fluorescein angiography (C). (D) Swept-source optical coherence tomography angiogram (SS-OCTA) of the preretinal vascular loop. (E) SS-OCT B-scan through vascular loop, as delineated by red dotted line in (D), with segmentation of the above SS-OCTA shown between the violet and blue line.
In 13.3% of cases with macular edema, OCTA clearly demonstrated areas of increased signal intensity that correlated well with pockets of intraretinal fluid on SD-OCT (Figure 4). In these cases, the fluid pockets were typically in the outer retina and demonstrated a relatively high reflectivity for a cystic space on SD-OCT. The fluid pockets were also typically associated with hyperreflective spots on OCT consistent with hard exudates, as well as yellow deposits on color photographs (Figure 4A). It should be noted that intraretinal pockets of hyporeflective fluid on SD-OCT (the more common variety of cystoid space seen in macular edema) were not seen on OCTA at all.
Swept-source optical coherence tomography angiogram (SS-OCTA) and spectral-domain OCT (SD-OCT) images of intraretinal fluid pockets with relatively high reflectivity. Conventional color fundus (A) and infrared image preview (B) of a subject with severe non-proliferative diabetic retinopathy with numerous hard exudates. SD-OCT scan (C) from bright green line in (B) through an area of intraretinal fluid demonstrates pockets of fluid in the mid and outer retina with relatively high reflectivity. No evidence of the intraretinal fluid is seen on fluorescein angiography (D), but it appears as a cloudy zone of high SS-OCTA signal in the mid-retina (F). Red arrows in (E–F) denote the location of the A-scan marked by the dotted red line in (C). A pocket of intraretinal fluid can also be seen inferior to the foveal avascular zone in the inner retina (E). (G) SS-OCT B scan through green dotted line in (E–F) with segmentation of the inner (blue to cyan), mid (cyan to red), and outer (red to yellow) retinal slices also showing pockets of fluid in the mid retina with relatively high reflectivity. (H) En face intensity SS-OCT scan of the full retina corresponding to the SS-OCTA scan in (E–F) does not demonstrate any artifactual decrease in OCT reflectivity.
Direct comparison of OCTA and fluorescein angiography found considerable variability in microaneurysm findings. Microaneurysms sometimes appeared on fluorescein angiography only, OCTA only, or both. In all cases, more microaneurysms were evident on fluorescein angiography than on OCTA. Careful comparison of the microaneurysms found on OCTA with fluorescein angiography showed a relatively poor concordance in both the shape and size of microaneurysms. Microaneurysms generally appeared smaller and in a variety of shapes, typically solid round (Figure 5F), round with dark centers (Figure 5J, bottom left), or fusiform (Figure 5J, upper left) on OCTA than in the corresponding fluorescein angiography, where they appeared as uniform round dots. In this particular case, 23% microaneurysms observed on fluorescein angiography were identified on OCTA.
Microaneurysm analysis in swept-source optical coherence tomography angiography (SS-OCTA) of diabetic subjects. (A) Fluorescein angiogram of a 67-year-old male with diabetic retinopathy with microaneurysms identified in small green squares. (B) Inner-retinal SS-OCTA with microaneurysms found on SS-OCTA identified with white circles and with microaneurysms identified on fluorescein angiography shown as green squares overlaid on the SS-OCTA image. (C) Mid-retinal SS-OCTA with microaneurysms found on SS-OCTA identified with white circles and with microaneurysms identified on fluorescein angiography shown as green squares overlaid on the image. (D–G) Magnified cyan inset from (A) shows microaneurysms seen and identified on fluorescein angiography (D–E) with the corresponding SS-OCTA of the inner (F) and mid (G) retina. Green squares in (E–G) indicate areas where microaneurysms were identified on fluorescein angiography. Magnified yellow inset from (A) shows microaneurysms seen and identified on fluorescein angiography (H–I) with the corresponding SS-OCTA of the inner (J) and mid (K) retina. Green squares in (I–K) indicate areas where microaneurysms were identified on fluorescein angiography. White arrowheads point out a microaneurysm seen only on SS-OCTA, but not clearly seen on fluorescein angiography.
Neovascularization of the Disc
One case of chronic, inactive neovascularization of the disc (NVD) was imaged on OCTA in a subject with quiescent proliferative DR (Figures 2I–J). NVD was confirmed on late phase fluorescein angiography. OCT angiograms closely correlated with these fluorescein angiographic findings. OCTA showed a complex of disorganized vessels protruding from the disc into the vitreous.
We used a SS-OCTA device using an intensity- and phase-based contrasting algorithm on subjects with DR. Depth-encoded, capillary-level resolution angiograms were generated that clearly demonstrated many of the major clinically significant vascular abnormalities found in DR. The qualitative agreement between fluorescein angiography and OCTA was greatest for areas of impaired capillary perfusion and least for intraretinal fluid and microaneurysms.
As in past diabetic studies with OCTA,6,8 there was less than a 1:1 correspondence between microaneurysms observed on OCTA and fluorescein angiography. Although currently an investigational technique, OCTA in combination with standard OCT imaging is at least as good as fluorescein angiography in the evaluation of the macular complications of diabetic retinopathy. It remains to be seen whether OCTA will alter the clinical management of DR or just serve as a complementary imaging method in place of fluorescein angiography.
It is important to recognize that OCTA assesses the flow of red blood cells in the retina as opposed to fluorescein angiography, which shows the flow of relatively small fluorescein molecules. In this context, the differences in microaneurysm appearances can be explained by several structural features of microaneurysms. From histopathology studies9,10 we know not all microaneurysms are patent and, therefore, may not have flow of red blood cells through them. On the other hand, the smaller fluorescein molecule may still diffuse into partially sclerosed microaneurysms making them visible on fluorescein angiography even though there is no red blood cell flow that can be detected by OCTA. In addition, adaptive optics scanning light ophthalmoscopy (AOSLO)11 has shown that even patent microaneurysms have different flow rates that may affect their appearance on OCTA.
It is also interesting to note that although microaneurysms on fluorescein angiography appear as hyperfluorescent dots, it is difficult to assess their true shape and size due to fluorescein leakage and background fluorescence. As opposed to uniform dots on fluorescein angiography, histological, SD-OCT, and AOSLO studies show that microaneurysms have a wide assortment of shapes, varying from saccular out-pouchings to irregular fusiform structures, and sizes ranging from 14 µm and 136 µm.10–12 One histological study of a cadaver eye with fluorescein angiography performed prior to enucleation noted that several microaneurysms’ appearance on fluorescein angiography were so out of proportion to their appearance on histology that they would not have been recognized histologically if not for the fluorescein angiogram.13 The difference could only be ascribed to fluoresce-in leakage and staining of the microaneurysm wall. Since OCTA does not detect leakage and staining, microaneurysms may appear both smaller and more irregular in shape on OCTA making them more difficult to readily detect. Both our results and findings by Ishibazawa et al.8 agree with this hypothesis as microaneurysms typically appear more saccular or fusiform in shape in our OCTA images. We also found that microaneurysms on OCTA typically appeared smaller than their fluorescein counterparts. It is encouraging that these results also are in agreement with histological studies of microaneurysm appearance.10,11,13
We consistently observed that areas of impaired capillary perfusion are more distinct on OCTA than corresponding fluorescein angiograms of “capillary non-perfusion.” The higher resolution of small vessels on OCTA also allowed observation of capillary loops on OCTA that were not visible clinically or on fluorescein angiography. There are a few likely reasons for these observations. Capillary-level resolution on fluorescein angiography is only possible in a very short transit (10- to 15-second) window, whereas fluorescein dye is predominantly in the capillaries.14 Beyond this interval, fluorescein extravasation into the tissue progressively decreases signal-to-noise from the capillaries and prevents visualization of the capillaries. One study of fluorescein angiography in 10 young normal subjects found that only about 30% of the fluorescein angiograms achieved adequate resolution for detailed capillary density analysis.15 This study also compared fluorescein angiograms to histological specimens of age-matched controls and found significantly lower capillary densities on fluorescein angiography compared to histology,15 suggesting fluorescein angiography is providing a limited picture of retinal capillaries. Lastly, it is possible that fluorescein dye flows through (or leaks into) capillaries that are not patent to the flow of red blood cells, thereby inaccurately portraying the area of true impaired perfusion.
Traditionally, vessels that extend into the preretinal space are thought to be exclusively neovascularization; however, our observations suggest that vascular loops can extend into the preretinal space without demonstrating the classical clinical findings of neovascularization. Classical neovascularization appeared distinctly different on our OCTA as a bundle of smaller disorganized vessels. It is interesting that in cases of untreated disc neovascularization imaged by Miura et al.16 and Ishibazawa et al.,8 the neovascularization appeared notably different than ours. Miura et al. used a Doppler OCT device in which the neovascularization appeared as tortuous medium sized vessels. Capillary-sized vessels were not seen well throughout the images, and it is possible that smaller vessels in the neovascularization were not seen due to the limitations of flow detection in Doppler OCT. Results from Ishibazawa et al. were quite different, with a much more lacy, indistinct appearance of the vessels than both Miura or our results, suggesting possible differences in the stages of neovascularization or variation in appearance due to type of imaging device and algorithm used.
The appearance of macular edema associated with unusually high reflectivity on SD-OCT was a unique non-vascular OCTA finding. Areas of non-vascular decorrelation signal have been noted before in regions with very fine internal structures. It has been hypothesized that very small eye motions or scanning changes may lead to OCTA signals from these small particles.17 In our cases of macular edema with relatively high reflectivity, it is possible that particulate debris (eg, constituents of hard exudates) within the edema could be generating OCTA signal. We have observed these intraretinal pockets of fluid with hyperreflectivity in cases of retinal vein occlusion, as well. In all cases, the pockets are associated with hard exudates on clinical exam and OCT B-scans. Therefore, we hypothesize that these hyperreflective pockets are a stage in the evolution of hard exudates from intraretinal fluid pockets.
OCTA has several inherent advantages for retinal angiography. Unlike fluorescein angiography, OCTA does not have side effects. Therefore, it can be performed as frequently as standard OCT scans and provide a more accurate assessment of blood flow changes than standard fluorescein angiography. This increased frequency of scanning has the potential to alter the clinical management of diabetic retinopathy, but this remains to be determined. Second, OCTA can consistently visualize capillary level changes in high resolution in both eyes repeatedly as opposed to fluorescein angiography, which can only image one eye during the transit phase of the angiogram. This will allow much more detailed evaluation of flow changes in both eyes of subjects with diabetic retinopathy. In addition, the ease and safety of OCTA may allow it to play an important role in assessing retinal vasculature early in the development of DR- a role fluorescein angiography has been unable to fill due to its contraindications.
In its current form, SS-OCTA does suffer from several drawbacks. It is relatively limited in field-of-view compared to fluorescein angiography. The typical field for SS-OCTA is 3 mm × 3 mm, whereas conventional fluorescein angiography can image 30° to 50° of the fundus and wide-field imaging methods can image up to 150°. SS-OCTA is also susceptible to motion artifacts, segmentation error, and any abnormalities that may block or disrupt conventional OCT signal. We believe many of these drawbacks offer substantial room for improvement as the technology progresses. Another potential drawback is the inability of SS-OCTA to detect leakage. Nevertheless, the absence of leakage on OCTA may also be useful in that there is no obscuration of capillary detail by the leakage.
- Matsunaga D, Yi J, Puliafito CA, et al. OCT angiography in healthy human subjects. Ophthalmic Surg Lasers Imaging Retina. 2014;45:510–515. doi:10.3928/23258160-20141118-04 [CrossRef]
- Spaide RF, Klancnik JM, Cooney MJ. Retinal vascular layers imaged by fluorescein angiography and optical coherence tomography angiography. JAMA Ophthalmol. 2015;133:45–50. doi:10.1001/jamaophthalmol.2014.3616 [CrossRef]
- An L, Shen TT, Wang RK. Using ultrahigh sensitive optical microangiography to achieve comprehensive depth resolved microvasculature mapping for human retina. J Biomed Opt. 2011;16:106013. doi:10.1117/1.3642638 [CrossRef]
- Jia Y, Tan O, Tokayer J, et al. Split-spectrum amplitude-decorrelation angiography with optical coherence tomography. Opt Express. 2012;20:4710–4725. doi:10.1364/OE.20.004710 [CrossRef]
- Mahmud MS, Cadotte DW, Vuong B, et al. Review of speckle and phase variance optical coherence tomography to visualize microvascular networks. J Biomed Opt. 2013;18:50901. doi:10.1117/1.JBO.18.5.050901 [CrossRef]
- Schwartz DM, Fingler J, Kim DY, et al. Phase-variance optical coherence tomography: a technique for noninvasive angiography. Ophthalmology. 2014;121:180–187. doi:10.1016/j.ophtha.2013.09.002 [CrossRef]
- Jia Y, Bailey ST, Hwang TS, et al. Quantitative optical coherence tomography angiography of vascular abnormalities in the living human eye. Proc Natl Acad Sci USA. 2015;112:E2395–2402. doi:10.1073/pnas.1500185112 [CrossRef]
- Ishibazawa A, Nagaoka T, Takahashi A, et al. Optical coherence tomography angiography in diabetic retinopathy: a prospective pilot study. Am J Ophthalmol. 2015;160:35–44.e1. doi:10.1016/j.ajo.2015.04.021 [CrossRef]
- Stitt AW, Gardiner TA, Archer DB. Histological and ultrastructural investigation of retinal microaneurysm development in diabetic patients. Br J Ophthalmol. 1995;79:362–367. doi:10.1136/bjo.79.4.362 [CrossRef]
- Foreman DM, Bagley S, Moore J, et al. Three dimensional analysis of the retinal vasculature using immunofluorescent staining and confocal laser scanning microscopy. Br J Ophthalmol. 1996;80:246–251. doi:10.1136/bjo.80.3.246 [CrossRef]
- Dubow M, Pinhas A, Shah N, et al. Classification of human retinal microaneurysms using adaptive optics scanning light ophthalmoscope fluorescein angiography. Invest Ophthalmol Vis Sci. 2014;55:1299–1309. doi:10.1167/iovs.13-13122 [CrossRef]
- Horii T, Murakami T, Nishijima K, et al. Optical coherence tomographic characteristics of microaneurysms in diabetic retinopathy. Am J Ophthalmol. 2010;150:840–848. doi:10.1016/j.ajo.2010.06.015 [CrossRef]
- De Venecia G, Davis M, Engerman R. Clinicopathologic correlations in diabetic retinopathy. I. Histology and fluorescein angiography of microaneurysms. Arch Ophthalmol. 1976;94:1766–1773. doi:10.1001/archopht.1976.03910040540013 [CrossRef]
- Johnson RN, Fu AD, McDonald HR, et al. Chapter 1: Fluorescein Angiography: Basic Principles and Interpretation. In: Ryan Stephen J SRS, Hinton David R., Schachat Andrew P., Wilkinson CP., Wiedemann Peter, eds. Retina (Fifth Edition). London: W.B. Saunders. 2013:2–50.e51. doi:10.1016/B978-1-4557-0737-9.00001-1 [CrossRef]
- Mendis KR, Balaratnasingam C, Yu P, et al. Correlation of histologic and clinical images to determine the diagnostic value of fluorescein angiography for studying retinal capillary detail. Invest Ophthalmol Vis Sci. 2010;51:5864–5869. doi:10.1167/iovs.10-5333 [CrossRef]
- Miura M, Hong YJ, Yasuno Y, et al. Three-dimensional vascular imaging of proliferative diabetic retinopathy by Doppler optical coherence tomography. Am J Ophthalmol. 2015;159:528–538.e523. doi:10.1016/j.ajo.2014.12.002 [CrossRef]
- Lumbroso B, Huang D, Jia Y, et al. Clinical Guide to Angio-OCT. 1st ed. New Delhi, India: Jaypee Brothers Medical Publishers; 2015.
|Number of subjects||18||6||9|
|Number of eyes||25||9||13|
|Male : Female||13 : 5||5 : 1||5 : 4|
|Age (mean +/− SD)||60.33 ± 10.72||62.17 ± 16.09||60.10 ± 10.91|
|Age range||34 – 74||41 – 80||37 – 77|
|Duration of disease (mean +/− SD, years)||13.75 ± 10.57a||9.80 ± 8.90b||19.13 ± 15.87b|
|Last-recorded HbA1c (mean +/− SD, %)||8.18 ± 1.62c||7.03 ± 1.19a||8.19 ± 1.42a|
|Time since last HbA1c (mean +/− SD, years)||0.54 ± 1.05c||0.28 ± 0.15a||0.61 ± 1.07a|
|Diabetic macular edema prevalence||0.12||0.89||0.36|