Ocular blood flow in preterm neonates

Retinopathy of prematurity (ROP) is a disorder affecting low birthweight, preterm neonates. In the preterm eye, the retina is not fully developed and neovascularization may occur at the margin between the developed vascular retina and undeveloped avascular retina. Without timely treatment by laser or intravitreal anti-vascular endothelial growth factor (VEGF) therapy, this can lead to tractional retinal detachment and blindness. Visualization of the retina in regular examinations by indirect ophthalmoscopy is hence the current standard of care, but the exams are stressful and interpretation of images is subjective. The upregulation of VEGF in ROP would suggest an increase in ocular blood flow. In this report, we evaluate the potential of ultrafast plane-wave Doppler ultrasound (PWU) to detect increased flow velocities in the orbital vessels supplying the eye in a gentle exam with objective findings. We imaged both eyes of 50 low-birthweight preterm neonates using 18 MHz PWU. Flow velocity in the central retinal artery (CRA) and vein (CRV), and the short posterior ciliary arteries were determined and values at each ROP Stage compared. We found significantly increased velocities in the CRA and CRV in Stage 3 ROP eyes, where intervention would be considered. We compared multivariate models for identifying Stage 3 eyes comprised solely of clinical factors, solely of Doppler parameters, and clinical plus Doppler parameters. The respective models provided areas under their respective ROC curves of 0.760, 0.812, and 0.904. PWU Doppler represents a gentle, objective means for identifying neonates at risk for ROP that could complement ophthalmoscopy.


Ocular blood flow in preterm neonates
Ronald H. Silverman 1,2* , Raksha Urs 1 , Jason D. Horowitz 1 , Osode Coki 1 & Leora Pinto 1 Retinopathy of prematurity (ROP) is a disorder affecting low birthweight, preterm neonates.In the preterm eye, the retina is not fully developed and neovascularization may occur at the margin between the developed vascular retina and undeveloped avascular retina.Without timely treatment by laser or intravitreal anti-vascular endothelial growth factor (VEGF) therapy, this can lead to tractional retinal detachment and blindness.Visualization of the retina in regular examinations by indirect ophthalmoscopy is hence the current standard of care, but the exams are stressful and interpretation of images is subjective.The upregulation of VEGF in ROP would suggest an increase in ocular blood flow.In this report, we evaluate the potential of ultrafast plane-wave Doppler ultrasound (PWU) to detect increased flow velocities in the orbital vessels supplying the eye in a gentle exam with objective findings.We imaged both eyes of 50 low-birthweight preterm neonates using 18  Retinopathy of prematurity (ROP) is a potentially devastating disorder of retinal vascular development occurring in low birthweight, preterm neonates.In spite of advances in neonatal intensive care, ROP remains a leading cause of pediatric blindness and potential lifelong visual impairment 1 .ROP incidence among premature infants in the United States has increased, especially among the Black and Hispanic populations 2 .
Starting at about 16 weeks gestation, the retina develops from the optic nerve outwards, a process that is complete at about 40 weeks 3 .The retinal vasculature is thus incompletely developed and immature in preterm neonates.
Oxygen plays a central role in retinal vessel development 4,5 .In utero, 'physiological hypoxia' allows normal retinal angiogenesis, primarily at the retinal periphery 6,7 , with VEGF playing a key role 8 .After preterm delivery and exposure to higher oxygen levels, there is cessation of vessel development and loss of already formed vessels.The avascular retina becomes more metabolically active and VEGF is upregulated; pathologic neovascularization with the potential for retinal detachment then follows 7 .
The current standard of care for preterm neonates entails examination of the eye by indirect ophthalmoscopy or wide-angle contact photography (RetCam, Natus Medical, Pleasanton, CA) on a regular basis.The first eye screening exam is appropriate at 4 weeks of age for neonates with gestational age (GA) of 27 weeks or more and at up to 9 weeks for progressively lower GA.Follow-up exams may occur weekly or less frequently, depending on ROP Stage and Zone observed in the prior exam 9 .ROP stages 10,11 and their characteristics are outlined in Table 1.
Expertise in performing indirect ophthalmoscopy, wide-angle digital imaging and image interpretation play a crucial role in ROP management, yet image interpretation is subjective.Examiners employing the latest RetCam imaging disagree 40% of the time as to even the presence or absence of ROP from its earliest stages to preplus and plus disease (i.e., abnormal retinal blood vessel tortuosity and dilation) [12][13][14] .There is thus a necessity for imaging methods offering objective, reproducible diagnostic criteria 11,12,15,16 .
Ophthalmoscopic examination of preterm neonates at risk for ROP has significant limitations beyond subjectivity: Exposure to mydriatic agents, bright light and use of a lid speculum and Flynn scleral depressor to conduct a detailed ophthalmoscopic or RetCam examination of the retina is stressful for the neonate, and not without potential morbidity [17][18][19][20][21] .
www.nature.com/scientificreports/Optical coherence tomography (OCT) and OCT-angiography (OCTA) have been explored for examination of the preterm eye.They are advantageous with respect to ophthalmoscopy in not requiring exposure to bright lights or scleral depressors and the exam can be performed via the non-dilated pupil (although dilation is advantageous) [22][23][24] .OCT, however, is limited to the posterior pole of the eye and provides only structural information.OCTA offers en face mapping of the retinal vasculature, and thus assessment of vessel tortuosity, density, and non-perfused areas (e.g., foveal avascular zone area).OCTA, however, is impeded due to eye motion during imaging (> 3 s) 25,26 .
Unlike ophthalmoscopy, ultrasound does not require mydriatics and unlike OCT can readily image the posterior structures of the eye even through a closed eyelid.Conventional B-scan, however, has relatively poor resolution compared to optical methods and offers no information on retinal vascular structure or blood flow.Hence, the use of B-scan ultrasound in ROP has been limited, with most reports citing its use for detection of retinal detachment [27][28][29][30] .Jokl et al. evaluated the potential of 10-20 MHz ultrasound B-scan to provide diagnostically significant information, reporting detection of the ridge between vascularized and avascular retina 31 and development of fibrovascular vitreous membranes portending retinal detachment 32 .
All of the above techniques image structure rather than function.The central role of VEGF upregulation in ROP, however, suggests that ocular blood flow might be increased in ROP.Doppler ultrasound has thus been used to study ocular blood flow, a functional property, in a number of ROP studies, with variable results.Some reports found no differences between ROP and non-ROP eyes 33 or in Stage 1-2 versus non-ROP 34 .Others reported significantly increased flow velocities in Stage-2 ROP 35 .
In the conventional Doppler systems utilized in the above studies, a focused ultrasound beam produced by a linear array is scanned to produce two-dimensional B-scan images.At each image position, the instrument must wait for echo data to be received before moving to the next position, which limits imaging speed to hundreds of B-scans per second.In pulsed Doppler, frequency shifts at specific ranges along the lines of sight are measured in real time to determine flow velocity, but with acoustic intensity exceeding FDA guidelines for ophthalmic imaging 36 .
Plane wave ultrasound (PWU) is an alternative approach that we utilized in a pilot study of 14 preterm neonates where we found increased arterial and venous flow velocities in Stage 3 ROP 37 .In PWU 38 , all the piezoelectric elements of a linear array act together to transmit an unfocussed wavefront.A focused image is created using a 'delay-and-sum' algorithm that brings echoes reaching each array element into phase by triangulating the time-of-flight from each image position.A B-scan image is thus formed with each transmit, allowing an imaging rate of over 20,000 B-scans/sec.Combining echo data from a series of angled transmits improves sensitivity and resolution 39 .By recording scans continuously over a cardiac cycle, Doppler analysis can be performed at any pixel position and systolic and diastolic flow velocities determined in post-processing.Furthermore, because the transmitted wavefront is unfocused, acoustic intensity is significantly less than would occur using a conventionally scanned, focused beam.PWU Doppler can be performed in compliance with ophthalmic regulatory standards, which is not the case in studies performed with conventional pulsed Doppler systems.This is a particularly important consideration in imaging the neonatal eye 40 .A representative PWU Doppler image and spectrograms depicting flow over the cardiac cycle is provided in Fig. 1.

Results
Table 2 provides a breakdown of the study population characteristics.ROP Stage at the time of exams is provided in Table 3.
Of the clinical factors, only birthweight (BW) and GA were found to differ significantly by Stage (p < 0.001 by ANOVA).Both factors were significantly different with respect to Stage 0 eyes at Stages 2 and 3 (p < 0.05) by post-hoc Dunnet's test.
The within exam standard deviations in the CRA were 12.2, 3.5 and 6.7 mmHg for the PSA, EDV and mean velocity respectively, corresponding to coefficients of variation of 0.28, 0.35 and 0.29.Variability in the CRV and SPCA were comparable.
ANOVA (Table 4) showed significant differences in Doppler flow velocity by Stage in the CRA and CRV.Flow velocities were significantly elevated in Stage 3.
Eleven eyes had preplus disease: 2 at Stage-1, 4 at Stage-2 and 5 at Stage-3.There were only 4 Zone-I (region surrounding the optic nerve head) eyes in the study population.Flow velocities were not significantly altered by presence of preplus or Zone-I disease, although this is uncertain given the small numbers of either preplus or Zone-I.Follow-up exams were performed in approximately 60% of subjects, with dropouts attributable to patient discharge.A noteworthy observation is that in Stage 3 ROP mean flow velocity in the CRA increased on followup by 8.1 ± 20.1 mm/s whereas, the change was only − 0.3 ± 9.9 mm/s in Stages 0-2.Although not statistically significant, this increase might be attributable to progression in Stage 3 eyes.www.nature.com/scientificreports/Clinical, Doppler and combined clinical + Doppler multivariate classification models are described in Table 6.Their ROC curves are plotted in Fig. 2 and the ROC area under curve for all models are shown in Table 6.

Discussion
The variability of individual measurements was significant, with coefficients of variability of around 30%.This was ameliorated by taking three scans per eye, reducing effective uncertainty of measurement of individual eyes by a factor of √3, or 1.73.
We found flow velocities in Stage 1-2 ROP to not differ significantly from Stage 0 eyes, which is consistent with prior studies using pulsed Doppler ultrasound 34,35 .We also confirmed our preliminary finding of significantly increased flow velocity at Stage 3 37 .
When comparing blood pressure in Stage 3 versus Stages 0-2, no measures reached statistical significance as univariate parameters, although diastolic blood pressure was borderline elevated (p = 0.056, two-tailed) in Stage 3 (53.3± 11.6 mmHg versus 47.4 ± 12.6 mmHg in Stages 0-2), and entered into the multivariate stepwise models.We have no hypothesis as to an underlying mechanism, and this remains to be verified in a larger study.
The ROC curves of the multivariate models indicate the tradeoff between sensitivity and specificity that would need to be considered in staging of ophthalmoscopy.We would ideally like a classifier offering 100% sensitivity and specificity for Stage 3 detection, but the best model only achieves perfect sensitivity at 50% specificity, which may represent an unacceptable false-positive rate.At 90% sensitivity, however, the combined model provides ~ 75% specificity, which may represent a practical tradeoff.There are several limitations to this study: Only 11 eyes were preplus and only 4 Zone-I; no patients were imaged pre-and post-treatment by laser or intravitreal anti-VEGF agents; Doppler measurements determined flow velocity, not volumetric flow; ophthalmoscopically determined ROP Stage and presence of preplus disease were treated as ground truth but are themselves subjective.
While the PWU Doppler technology utilized in this study utilized customized software and was implemented on a research ultrasound platform, turnkey PWU systems are now becoming available (albeit not configured for ophthalmology).It is probable that systems suitable for pediatric ophthalmology will in the future offer safe and gentle assessment of ocular blood flow so that these findings could impact ROP management.
In conclusion, PWU could be used to reduce the number of stressful ROP screening dilated ophthalmoscopic exams that babies would have to undergo.PWU could be combined with additional new screening algorithms to implement even further reduction.Possibly the blood flow measurements could contribute to prognostication and refinement of indications for treatment in combination with ophthalmoscopy and/or photography.

Methods
This research followed the tenets of the Declaration of Helsinki and was approved by the Columbia University Institutional Review Board as well as by a separate neonatal intensive care unit (NICU) research ethics and safety committee.Informed consent was obtained from the parents after explanation of the nature, risks, and benefits of the study.
Neonate subjects were inpatients in the New York-Presbyterian/Morgan Stanley Children's Hospital NICU.
MHz PWU.Flow velocity in the central retinal artery (CRA) and vein (CRV), and the short posterior ciliary arteries were determined and values at each ROP Stage compared.We found significantly increased velocities in the CRA and CRV in Stage 3 ROP eyes, where intervention would be considered.We compared multivariate models for identifying Stage 3 eyes comprised solely of clinical factors, solely of Doppler parameters, and clinical plus Doppler parameters.The respective models provided areas under their respective ROC curves of 0.760, 0.812, and 0.904.PWU Doppler represents a gentle, objective means for identifying neonates at risk for ROP that could complement ophthalmoscopy.

Figure 1 .
Figure 1.Left: Representative plane-wave color flow Doppler image of neonatal posterior pole.Right: Spectrograms of the CRA, CRV and SPCA.

Figure 2 .
Figure 2. ROC curves for the clinical Model 1, Doppler Model 2, and Doppler plus clinical Model 3.

Table 1 .
ROP Stages and their characteristics.Followup between exams ranged from 14 to 27 days, averaging 21 days.Between initial and followup exams, eight eyes progressed from Stage 1 to Stage 2 and one from Stage 2 to Stage 3. Flow velocity in the CRA and SPCA increased in this group (e.g., by 7.1 ± 15.7 mm/s PSV in the CRA), but given the small number of cases, this was not statistically significant.Because flow velocities were comparable at Stages 0-2 and increased only at Stage 3, we compared Stage 3 with combined Stages 0-2.ANOVA results comparing flow in Stage 3 versus Stages 0-2 in both the initial exam and follow-up exam are presented in Table5.

Table 3 .
ROP Stage at initial and follow-up exams.

Table 4 .
ANOVA of flow velocity by Stage for each vessel.*Indicates p ≤ .05 and **indicates p ≤ 0.01 by posthoc Dunnett's test with respect to Stage 0.

Table 5 .
One-way ANOVA for Stage 3 versus Stages 0-2 for all exams, initial exam only and follow-up exams only.Since treatment is not normally performed prior to Stage 3, these findings are potentially significant in terms of management: if Stage 3 could be detected by PWU, confirmation by ophthalmoscopy would be prioritized and conversely, if PWU does not indicate Stage 3 ROP, stressful ophthalmoscopic examinations could be performed less frequently.

Table 6 .
Discriminant analysis models including only clinical parameters (Model 1), only Doppler parameters (Model 2) and Doppler plus clinical parameters (Model 3).Standardized coefficients represent the weight of each parameter in models.