Optical coherence tomography angiography (OCTA) allows visualization of ocular vessels by detecting motion contrast from flowing blood. Motion contrast can be detected by two different validated methods: the “speckle decorrelation,” which detects intensity changes in OCT structural image, and the “phase variance,” which assesses changes in the phase of light wave.1–7
OCT samples a discrete tissue volume and generates a numerical value based on the reflectivity of the volume. This numerical value corresponds to a voxel (that is a portmanteau of “volume” and “pixel”) in the image displayed. Each voxel is an estimation of the actual reflective properties of a small volume of tissue. OCTA repeatedly scans the same discrete tissue volume and detects proportional changes in reflectivity signal. If the changes are above a selected level, or threshold, flow is considered to be present.8 In other words, OCTA creates images by scanning the same region repeatedly. This allows for detection of changes in reflectance due to blood flow in vessels as compared to stationary tissue structures that does not change between repeated scanning. All these data are processed and displayed as an image that resembles an angiogram, similar to fluorescein angiography or indocyanine green angiography, but dye-less.9
However, unlike other imaging systems, OCTA produces artifacts, that are extraneous image information. They might be due to image acquisition, intrinsic characteristics of the eye or ocular pathology, eye motion, image processing, or display strategies.8
Many types of artifacts exist, but the most common is due to eye movement. Such findings are called motion artifacts and may occur both in axial and transverse directions. However, the major source of artifacts in OCTA is transverse eye motion.8 A momentary transverse change in location of the scan can be caused by a microsaccade with return of fixation to the original position. In that time, the different B-scans do not match because they correspond to different areas of the tissue. The images acquired during such movements will show large change for one or more adjacent B-scans, resulting in white line artifacts in the OCTA image. Furthermore, a temporary shift in fixation can produce an artifact where there is a lateral shearing of the image, or it can create a shift in which areas of the fundus are lost.8
In order to reduce the number of artifacts per image, particularly transverse motion artifacts, Optovue (Freemont, CA) has used different strategies. Initially, the company introduced a software-based method10 by which OCTA volumes of the same retinal area are acquired repeatedly with horizontal and then vertical scans; more recently, the company developed an eye-tracking (ET) technology called DualTrac Motion Correction Technology (MCT). However, such strategies may imply that longer examinations are required compared to OCT or OCTA without ET.
In this study, our aim was to compare the performances of the spectral-domain OCTA (SD-OCTA) Avanti with AngioVue imaging system with or without ET by evaluating the execution time needed to perform the complete examination, the number of transvers motion artifacts in each image, and number of images available for clinical analysis.
Patients and Methods
Patients affected by different retinal diseases and healthy subjects presenting between November 2016 and December 2016 at the Medical Retina & Imaging Unit of the Department of Ophthalmology, University Vita-Salute San Raffaele in Milan, Italy, were enrolled in the study. All patients and healthy subjects underwent two consecutive OCTA examinations (Exam A and B) in both eyes using the same device. After the first examination (Exam A), the operator inactivated the ET technology. The study was conducted in agreement with the Declaration of Helsinki for research involving human subjects and was approved by the local institutional review board. Included patients signed a written consent to participate to the study.
Inclusion criteria for the study group were age older than 18 years, sufficiently clear ocular media, and adequate pupillary dilation and fixation to permit execution of OCTA imaging. Excluded criteria were patients and healthy subjects refusing to participate to the study.
The OCTA used for image acquisition was the AngioVue imaging system, installed on the SD-OCT Avanti Widefield. Scanning pattern for en face angiographic visualization included a 3 mm × 3 mm and a 6 mm × 6 mm scan pattern on the retina. Two orthogonal OCTA volumes are acquired for orthogonal registration using motion correlation technology to minimize motion artifacts. The split-spectrum amplitude-decorrelation angiography (SSADA) algorithm is used to extract the OCTA information. SSADA 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 blood pulsation.11 Even if motion correction technology and SSADA algorithm have reduced OCTA image artifacts, Optovue has recently added DualTrac Motion Correction Technology to AngioVue OCTA in order to improve image quality. DualTrac consists of post-processing motion correction technology and real-time ET technology.
Expert operators acquired four images for each subject (a 3 mm × 3 mm image and a 6 mm × 6 mm image for each eye) and measured the execution time needed to complete and save the exam. After Exam A, the operators inactivated the ET and repeated all the procedures described above for Exam B. From this point on, “Exam A” will indicate the examination with ET, whereas “Exam B” will refer to the examination without ET. Operators were not allowed to repeat the acquisition in the presence of low-quality examination or scarce collaboration by the subjects studied. Execution time was measured with a standard chronometer (iPhone 6, iOS 9.0; Apple, Cupertino, CA). The measurement started at the beginning of the first scan and stopped after saving the last scan, thus excluding preparation of the machine, positioning and registration of patients. Then, two readers (LADV, RS) independently counted the number of artifacts per image. In the analysis of artifacts, the readers excluded images with signal strength lower than 50 out of 100 and low-quality images impossible to analyze. When the image analysis was excluded for one examination, the corresponding image obtained during another examination was excluded, as well.
Statistical calculations were performed using Prism version 6.0h (GraphPad Software, La Jolla, CA). The difference between execution time and the number of motion artifacts counted in images obtained during the two examinations was generated by conducting Wilcoxon matched pairs signed-rank test. The difference of execution time and number of artifacts per image between patients and healthy subjects was calculated by conducting a Mann-Whitney test (unpaired, nonparametric). Nonparametric Spearman correlation between the numbers of artifacts measured by the two readers was also calculated. Additionally, correlation analyses with the nonparametric Spearman correlation were calculated between number of artifacts, and signal strength, execution time or patients' age for both devices. Differences between subtypes of retinal disease were evaluated by using analysis of variance and Bonferroni post hoc tests. The chosen level of statistical significance was a P value less than .05.
A total of 100 eyes of 50 subjects (23 male and 27 female; mean age: 57.5 years ± 20.1 years) were included in the analysis (Table 1). Seven (five male and two female; mean age: 27.3 years ± 10.3 years) out of 50 subjects were healthy subjects, whereas 43 (18 male and 25 female; mean age: 62.4 years ± 16.8 years) out of 50 subjects were affected by different retinal diseases in one or both eyes. Twenty-three subjects were affected by age-related macular degeneration (AMD), 11 by central serous chorioretinopathy (CSC), and nine by diabetic retinopathy (DR) (Table 2). Total acquired images were 400:200 images during Exam A and 200 during Exam B (Table 1).
Demographics and Results
Results Divided by Subtypes of Retinal Disease
Mean execution time was 224.9 seconds ± 67.7 seconds for Exam A, which was statistically significantly higher compared to the execution time for Exam B (181.3 seconds ± 27.2 seconds; P < .0001) (Figure 1 and Table 1).
Mean execution time required by the examination with and without eye-tracking in the overall population and in the two groups separately (patients and healthy subjects).
If we consider the Exams A and B separately, there are not differences in term of execution time between patients and healthy subjects (P = .09 and P = .27, respectively).
On the contrary, if we analyze patients' and healthy subjects' examinations separately, execution time for Exam A is shorter compared to Exam B (183.0 seconds ± 28.2 seconds vs. 231.1 seconds ± 70.5 seconds; P < .0001 for patients' examinations) (170.6 seconds ± 17.5 seconds vs. 187.0 seconds ± 27.3 seconds; P = .03 for healthy subjects' examination) (Figure 1; Table 1).
We also compared the mean execution between subtypes of retinal diseases (Table 2). Of note, in both Exams A and B, the mean execution time was statistically significantly higher in patients with AMD compared with patients with CSC (P < .001 with ET; P = .01 without ET) and patients with DR (P < .001 with and without ET). However, in all the retinal disease groups the mean execution time for Exam A was significantly higher when compared with Exam B (P < .001).
Analysis of Image Quality
There were 400 images acquired initially; however, after exclusion, 218 images were available for analysis (109 acquired during Exam A, 109 acquired during Exam B).
Overall, 124 images were excluded because maximum signal strength was below 50/100, or because of low-quality images impossible to analyze. Fifty-eight images were excluded because the corresponding image obtained during the other examination was excluded, as well.
In the overall population, 45 out of 200 images (22.5%) were excluded because maximum signal strength was below 50/100 for Exam A, and 58 out of 200 (29.0%) were excluded for Exam B. The readers also excluded five of 200 images (2.5%) obtained using ET and 16 of 200 images (8.0%) obtained without using ET because of low quality impossible to analyze (Figure 2; Table 1).
With eye-tracking (ET): Percentage of available images, low signal strength images, and images impossible to analyze obtained using ET (75.0%, 22.5%, and 2.5%, respectively) in the overall population. Without ET: percentage of available images, low signal strength images, and images impossible to analyze obtained without using eye-tracking (63.0%, 29.0%, and 8.0%, respectively) in the overall population.
Within the group of healthy subjects, 56 images were obtained (28 during Exam A, 28 during Exam B), but after exclusion conducted by readers, 46 images were available for artifact analysis (23 for Exam A, 23 for Exam B). For Exam A, readers excluded one image because maximum signal strength was below 50/100 and one image because it was impossible to analyze. For Exam B, readers also excluded four images: two because maximum signal strength was below 50/100 and two because of low quality impossible to analyze (Table 1).
Within the group of patients, a total of 344 images were acquired (172 during Exam A and 172 during Exam B). For Exam A, readers excluded 44 images and four images, respectively, because of low signal strength and low quality. For Exam B, 56 images were excluded because of low signal strength and 14 images because of low quality. Ultimately, 172 images were available for artifact analysis were (Table 1).
It is noteworthy that during Exam A, it was possible to obtain a higher number of images available for the analysis of artifacts compared to Exam B (150 [75% of total images] vs. 126 [63% of total images]) (Figure 2). Nevertheless, within the group of healthy subjects, the difference between Exams A and B in term of images available for artifacts analysis was less evident (25 [89%] for Exam A, 23 [82%] for Exam B).
Analysis of Motion Artifacts
The mean number of artifacts (Figure 3) was significantly lower in images obtained during Exam A compared to those obtained during Exam B (4.9 ± 2.8 vs. 9.0 ± 4.2; P < .0001) (Figure 4). Considering both the control and patient groups separately, the mean number of artifacts continued to be significantly lower in Exam A compared to Exam B (5.4 ± 3.0 vs. 8.5 ± 2.5; P < .0001 for the group of healthy subjects; 5.1 ± 3.1 vs. 9.0 ± 4.5; P < .0001 for the group of patients).
Examples of artifacts with (a, c, e, g) and without eye-tracking (b, d, f, h), in a healthy subject (a, b c, d) and a patient with age-related macular degeneration (e, f, g, h). Red lines highlight the motion artifacts.
Number of motion artifacts per image. AngioVue with eye-tracking (ET) versus AngioVue without ET in the overall population and in the two groups separately (patients and healthy subjects).
The difference in the mean number of artifacts between patients and healthy subjects was not statistically significant neither for Exam A (P = .6) or Exam B (P = .6).
The number of artifacts per image counted by the two readers were strongly correlated (with ET, r = 0.8, P < .0001; without ET, r = 0.8, P < .0009).
There was no correlation between the number of artifacts and execution time (P = .2 with ET, P = .8 without ET) or patients' age (P = .2 with ET, P = .2 without ET). There was a weak inverse correlation between the number of motion artifacts and signal strength (r= −0.4, P < .0001 with ET; r= −0.3, P < .0001 without ET).
When we compared the mean number of artifacts between subtypes of retinal diseases (Table 2), the only statistically significant difference we found for Exam A was the one between the DR and CSC groups (P = .005). The mean number of artifacts for Exam B did not vary between subtypes of retinal disease (P = .603). In all the retinal disease groups, the mean number of motion artifacts for Exam A was significantly lower when compared with Exam B (P < .001).
In this study we evaluated the performance of real-time ET technology DualTrac Motion Correction Technology in terms of execution time needed to complete an examination, number of images available for clinical analysis, and number of motion artifacts per image.
Execution time refers to the time that expert operators needed to acquire four images per subject. According to our results, a complete examination executed with ET requires a longer time of around 45 seconds compared to the examination without ET (P < .0001). Of note, the absence of retinal pathology showed a trend toward reduction in execution time, even though not statistically significant (P = .09 for Exam A; P = .27 for Exam B). Nevertheless, considering the groups of patients and healthy subjects separately, mean execution time is significantly shorter for the examination without ET compared to the examination with ET. In addition, the mean execution time is higher for Exam A than Exam B even if subtypes of retinal diseases are separately evaluated.
Finally, AMD patients require a longer examination time when compared to other patients. This evidence could be explained by the poor fixation of AMD patients.
The other goal of this study was to count the number of motion artifacts per image after acquisition with and without ET technology. It is noteworthy that a large number of scans were not included in the analysis and thus, after exclusion, only 218 images (55% of total acquired images) were available for analysis. A higher number of images available for analysis were obtained using ET (150 [75%]) compared to examination without ET (126 [63%]). Moreover, similar results were found in the group of patients, as 72% of total images taken during examination with ET were available for analysis, but only the 59% during examination without ET. In the group of healthy subjects, both Exams A and B produced an elevated number of available images (26 [93%] and 24 [86%], respectively). These data suggest that ET does not improve the percentage of available images for healthy subjects possibly because of good fixation, whereas in patients, it definitely improves the number of available images.
Artifacts, which affect every imaging system, produce misleading image information, and in OCTA they can be caused by several factors. Motion artifacts (ie, artifacts derived from eye movement) are common in OCTA.8 Our results showed that the mean number of motion artifacts per image is significantly lower in images obtained using ET (4.9 artifacts per image ± 2.8 artifacts) compared to images obtained during examination without ET (9.0 artifacts per image ± 4.2 artifacts). Furthermore, the use of ET reduces the number of motion artifacts, regardless of retinal diseases.
Interestingly, in the current study, images from healthy subjects did not present a significant lower number of artifacts compared to images from patients (as it was expected due to poor fixation in diseased eyes) in both types of examination.
Unexpectedly, when the mean number of motion artifacts for Exam A was compared between subtypes of retinal diseases, patients with CSC had a lower number of motion artifacts than patients with DR (P = .005). On the contrary, the same analysis for Exam B did not disclose any statistically significant difference.
The different number of motion artifacts is strictly linked to the presence of retinal ET, primarily developed to reduce motion artifacts. On the other hand, retinal ET requires a longer time in order to acquire and elaborate images, and thus an overall longer execution time.
Although our results seem to attribute many advantages to the use of ET, further studies are needed to overcome the limits of this study, including analysis of other types of artifacts, a larger number of subjects, which may clarify other aspects of OCTA image quality.
Nowadays, OCTA is becoming a fundamental ophthalmologic clinical practice tool, as demonstrated by numerous applications in several retinal diseases.12–22 For these reasons, and considering the burden of patients in every day clinical practice, OCTA devices need to provide reliable and good-quality exams in a very short time; nevertheless, despite the longer examination time, the advantages in term of image quality and availability justify the use of eye-tracking technology.
- Fingler J, Readhead C, Schwartz DM, Fraser SE. Phase-contrast OCT imaging of transverse flows in the mouse retina and choroid. Invest Ophthalmol Vis Sci. 2008;49(11):5055–5059. doi:10.1167/iovs.07-1627 [CrossRef]
- Fingler J, Schwartz D, Yang C, Fraser SE. Mobility and transverse flow visualization using phase variance contrast with spectral domain optical coherence tomography. Opt Express. 2007;15(20):12636–12653. doi:10.1364/OE.15.012636 [CrossRef]
- Mariampillai A, Standish BA, Moriyama EH, et al. Speckle variance detection of microvasculature using swept-source optical coherence tomography. Opt Lett. 2008;33(13):1530–1532. doi:10.1364/OL.33.001530 [CrossRef]
- Mariampillai A, Leung MK, Jarvi M, et al. Optimized speckle variance OCT imaging of microvasculature. Opt Lett. 2010;35(8):1257–1259. doi:10.1364/OL.35.001257 [CrossRef]
- Kim DY, Fingler J, Werner JS, Schwartz DM, Fraser SE, Zawadzki RJ. In vivo volumetric imaging of human retinal circulation with phase-variance optical coherence tomography. Biomed Opt Express. 2011;2(6):1504–1513. doi:10.1364/BOE.2.001504 [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(10):106013. doi:10.1117/1.3642638 [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(5):50901. doi:10.1117/1.JBO.18.5.050901 [CrossRef]
- Spaide RF, Fujimoto JG, Waheed NK. Image artifacts in optical coherence tomography angiography. Retina. 2015;35(11):2163–2180. doi:10.1097/IAE.0000000000000765 [CrossRef]
- Spaide RF, Klancnik JM Jr., Cooney MJ. Retinal vascular layers imaged by fluorescein angiography and optical coherence tomography angiography. JAMA Ophthalmol. 2015;133(1):45–50. doi:10.1001/jamaophthalmol.2014.3616 [CrossRef]
- Jia Y, Tan O, Tokayer J, et al. Split-spectrum amplitude-decorrelation angiography with optical coherence tomography. Opt Express. 2012;20(4):4710–4725. doi:10.1364/OE.20.004710 [CrossRef]
- Huang D, Jia Y, Gao SS, Lumbroso B, Rispoli M. Optical coherence tomography angiography using the Optovue device. Dev Ophthalmol. 2016;56:6–12. doi:10.1159/000442770 [CrossRef]
- Bandello F, Corbelli E, Carnevali A, Pierro L, Querques G. Optical coherence tomography angiography of diabetic retinopathy. Dev Ophthalmol. 2016;56:107–112. doi:10.1159/000442801 [CrossRef]
- Iafe NA, Phasukkijwatana N, Sarraf D. Optical coherence tomography angiography of type 1 neovascularization in age-related macular degeneration. Dev Ophthalmol. 2016;56:45–51. doi:10.1159/000442776 [CrossRef]
- Querques G, Miere A, Souied EH. Optical coherence tomography angiography features of type 3 neovascularization in age-related macular degeneration. Dev Ophthalmol. 2016;56:57–61. doi:10.1159/000442779 [CrossRef]
- Battaglia Parodi M, Pierro L, Gagliardi M, Lattanzio R, Querques G, Bandello F. Optical coherence tomography angiography in dystrophies. Dev Ophthalmol. 2016;56:159–165. doi:10.1159/000442808 [CrossRef]
- Querques G, Corvi F, Querques L, Souied EH, Bandello F. Optical coherence tomography angiography of choroidal neovascularization secondary to pathologic myopia. Dev Ophthalmol. 2016;56:101–106. doi:10.1159/000442800 [CrossRef]
- Waheed NK, Moult EM, Fujimoto JG, Rosenfeld PJ. Optical coherence tomography angiography of dry age-related macular degeneration. Dev Ophthalmol. 2016;56:91–100. doi:10.1159/000442784 [CrossRef]
- Souied EH, Miere A, Cohen SY, Semoun O, Querques G. Optical coherence tomography angiography of fibrosis in age-related macular degeneration. Dev Ophthalmol. 2016;56:86–90. doi:10.1159/000442783 [CrossRef]
- Srour M, Querques G, Souied EH. Optical coherence tomography angiography of idiopathic polypoidal choroidal vasculopathy. Dev Ophthalmol. 2016;56:71–76. doi:10.1159/000442781 [CrossRef]
- Pierro L, Battaglia Parodi M, Rabiolo A, Introini U, Querques G, Bandello F. Optical coherence tomography angiography of miscellaneous retinal disease. Dev Ophthalmol. 2016;56:174–180. doi:10.1159/000442810 [CrossRef]
- Baumal CR. Optical coherence tomography angiography of retinal artery occlusion. Dev Ophthalmol. 2016;56:122–131. doi:10.1159/000442803 [CrossRef]
- Novais EA, Waheed NK. Optical coherence tomography angiography of retinal vein occlusion. Dev Ophthalmol. 2016;56:132–138. doi:10.1159/000442805 [CrossRef]
Demographics and Results
|Overall Population||Patients||Healthy Subjects|
|Number of Subjects Enrolled||50||43||7|
|Mean Age (Years ± SD)||57.5 ± 20.1||62.4 ± 16.8||27.3 ± 10.3|
|Total Images Acquired (With and Without ET)||400||344||56|
|Images Available for Analysis of ArtifactsExam A (With ET)||218||172||46|
|Total Images Acquired||200||172||28|
| Available images (% on total images)||150 (75.0%)||124 (72.1%)||26 (92.9%)|
| Low signal strength (% on total images)||45 (22.5%)||44 (25.6%)||1 (3.6%)|
| Images impossible to analyze (% on total images)||5 (2.5%)||4 (2.3%)||1 (3.6%)|
|Mean Execution Time (Seconds ± SD)||224.9 ± 67.7||231.1 ± 69.9||187 ± 25.7|
|Mean Number of Motion Artifacts (Number ± SD)Exam B (Without ET)||4.9 ± 2.8||5.1 ± 3.1||5.4 ± 3.0|
|Total Images Acquired Without ET||200||172||28|
| Available images (% on total images)||126 (63%)||102 (59.3%)||24 (85.7%)|
| Low signal strength (% on total images)||58 (29.0%)||56 (32.6%)||2 (7.1%)|
| Images impossible to analyze (% on total images)||16 (8.0%)||14 (8.1%)||2 (7.1%)|
|Mean Execution Time (Seconds ± SD)||181.3 ± 27.2||183.0 ± 28.0||170.6 ± 16.5|
|Mean Number of Motion Artifacts (Mean ± SD)||9.0 ± 4.2||9.0 ± 4.5||8.5 ± 2.5|
Results Divided by Subtypes of Retinal Disease
|Number of Subjects Enrolled||23||11||9|
|Mean Execution Time w/ ET, Seconds (Mean ± SD)||260.00 ± 81.06||206.89 ± 35.40||193.67 ± 23.03|
|Mean Execution Time w/o ET, Seconds (Mean ± SD)||192.48 ± 32.43||176.11 ± 19.46||171.44 ± 10.50|
|Mean Number of Artifacts w/ ET (Mean ± SD)||5.31 ± 3.14||3.88 ± 2.34||6.84 ± 3.32|
|Mean Number of Artifacts w/o ET (Mean ± SD)||9.58 ± 3.88||8.34 ± 5.32||9.47 ± 3.20|