Reliability of Confocal Choroidal Laser Doppler Flowmetry

Abstract

purpose. To assess the variability in choroidal laser Doppler flowmetry (LDF).

methods. Five measurements of choroidal blood flow were obtained on 5 days in both eyes of 10 subjects. Reliability, sensitivity, and sample size calculations were performed. Yield, the intensity of the returning light (direct current [DC]) divided by the signal amplification (gain) used during recording, was calculated (yield = DC/gain). The correlation between yield and the LDF parameters velocity, volume, and flux was analyzed. Reliability, sensitivity, and sample size calculations were repeated after correcting for this relationship. The influence of different light-scattering properties on the returning signal was assessed in a model eye.

results. Yield and the LDF parameters, velocity, volume, and flux, correlated significantly in a regression model. After the influence of yield was partialized out, reliability, reproducibility, sensitivity, and statistical power improved markedly. The high reliability of yield in a given eye (87%) suggests, that, besides varying recording settings on different days, individual specific scattering properties within the eye influence LDF measurements. Comparison of model eye data with human data disclosed a low signal-to-noise ratio with decreasing yield. Correcting for yield did not affect the sensitivity to blood flow changes induced by hypercapnia or a suction cup.

conclusions. Tissue-scattering properties alter the characteristics of the returning light, and recording settings affect the outcome of the analysis of the returning light during LDF measurements. Partializing out the influence of yield markedly improves the reproducibility of choroidal LDF.

The choroid plays an important role in the supply of nutrients to and removal of waste products from the outer region of the fovea. Abnormal choroidal blood flow may disrupt the normal retinal function and lead to alterations in visual functions. Several methods have been developed for evaluating blood flow in the choroid. Destructive or invasive methods ^{1 2 3 4 5 6 7 8} are restricted to experimental studies in animals. Methods applicable to humans are limited to estimates of the pulsatile component of choroidal blood flow ^{9 10} or are based on angiographic techniques ^{11 12} or laser technology. ^{13 14}  

Laser Doppler flowmetry (LDF) has repeatedly been applied for the assessment of choroidal blood flow in humans. ^{14 15 16 17 18 19} The marked variability of measurements obtained with this technology precludes a wide application of this methodology. In a preparation devoid of non–blood-flow–related movement, this variation is thought to be due to variation in vascular density and vessel orientation within the relatively small volume of tissue sampled by LDF. In the present study, we investigated the sources of the variation during choroidal LDF, using a new compact, confocal choroidal laser Doppler flowmeter (Institut de Recherche en Ophtalmologie, Sion, Switzerland). ¹⁵ This instrument was used in the present study, because, theoretically, it comes close to the assumption that the site of LDF measurement does not change between measurements. 

Subjects and Methods

Compact Choroidal LDF

In choroidal LDF, laser light scattered by moving red blood cells is shifted in frequency by amounts proportional to the speed of the red blood cells. The returning signal is analyzed according to the procedure developed by Bonner and Nossal ²⁰ and the blood flow parameters velocity, volume, and flux are computed. Velocity is expressed in kilohertz and volume and flux are expressed in arbitrary units (AU). 

The optical system for the delivery of the laser beam and the detection of the scattered light of the new compact, confocal choroidal laser Doppler flowmeter (Institut de Recherche en Ophtalmologie) is based on a confocal arrangement and has been described in detail elsewhere. ¹⁵ Briefly, a polarized laser source (λ = 785 nm, 100 μm) is relayed with a 1:1 optical system (laser beam at the cornea: width, 1.3 mm; power, 90 μW) and focused at the subject’s retina (spot in the retinal image plane, 10–20 μm in diameter; optical thickness of confocal layer, 300 μm). The point laser source, the point of illumination of the fovea, and the detecting optical fiber are located in conjugated planes. The scattered light is collected by an optical system organized with six fibers arranged circularly around the central fixation point along a circle of diameter of 180 μm (within the avascular zone of the fovea). 

Subjects

Ten healthy nonsmoking volunteers (seven women, three men; age, 24–38 years; mean ± SD, 33.17 ± 3.9) were recruited. The procedures were approved by the local ethics committee, the tenets of Declaration of Helsinki were observed, and each subject signed an informed consent form before undergoing examination. Included were individuals with no history of ocular or systemic disease, no history of long-term or current use of systemically active or topical medication, and no history of drug or alcohol abuse. Further inclusion criteria were normal systolic (100–140 mm Hg) and diastolic (60–90 mm Hg) blood pressure, best corrected visual acuity of more than 20/25 in both eyes, ametropia within −3 to + 3 D of spherical equivalent and less than 1 D astigmatism in both eyes, pupil diameter of at least 4 mm, applanatory intraocular pressure (IOP) below 20 mm Hg in both eyes, and no pathologic findings in slit lamp examination and indirect fundoscopy. 

Measurement of the Blood Flow Parameters

Subjects were examined after an overnight fast and were asked to refrain from alcohol and caffeine for 12 hours before the trial days. A resting period of at least 30 minutes was scheduled for each subject. Stable baseline conditions were established and were ensured by repeated measurements of blood pressure. The subjects were seated with the head stabilized in a slit lamp headrest. Care was taken to standardize the subject’s head position by aligning marks on the headrest with anatomic landmarks on the forehead, chin, and temporal orbital rim. The subjects were asked to fixate the red-light spot within the ocular and to adjust the focus by turning the ocular until the smallest possible size of the red light spot was obtained. The ocular-to-cornea distance was set between 1.5 and 2 cm and held constant in all the subsequent recordings. In addition, constant, very low-level artificial room illumination was used throughout all the experiments. Twenty- to 25-second recordings were obtained in each eye. A stable direct current (DC) during a recording was used as a criterion for proper fixation. All subjects were perfectly familiar with the measurement technique. 

Reproducibility of the Measurements

Five measurements of choroidal blood flow were obtained in both eyes of 10 subjects on five consecutive days. All the measurements were performed at the same time of day in each subject (±15 minutes). Right and left eyes were measured in random order. Once a recording was obtained and saved, no further analysis was attempted until the end of data collection in all the scheduled examinations. Consequently, the examiner (KG) was masked to the previous measurement results of a given subject. The intraindividual (repeated measurements in the same eye) coefficients of variation (CV = 100 × SD/mean) were assessed. The reliability (R, intraclass correlation coefficient) was assessed according to the formula ²¹  

\[R{=}100{\times}(\mathrm{Var}(B_{i})/(\mathrm{Var}(B_{i}){+}\mathrm{Var}(e_{ij}))\]

where Var (B _i ) is the variance due to the biological component (one randomly selected eye in 10 subjects) and Var(e _ij ) is the variance due to the error (five repeated measurements in the same eye) component. 

Detection Sensitivity of the Blood Flow Parameters

The minimum statistically significant change (S) that can be detected in a group of 10 subjects was calculated according to the formula ²¹  

\[S{=}(t{\times}SD)/({\surd}10{\times}P_{\mathrm{mean}}){\times}100\]

where P _mean is the mean value of all measurements, SD the SD of the difference between paired measurement for all subjects, and t the two-tailed value of the t distribution at a 0.05 significance level for 9 df. ²² A range of sensitivity values was calculated by comparing the first recordings with the series that followed. 

Sample Size Calculations

Sample size calculations for a change of 5%, 10%, 15%, and 25% from baseline were performed for the parameter flux, with the α error set to 0.05 and the power (1-β) set to 0.8. A range of sample size calculations was calculated by comparing the first recordings to the series that followed. 

Correlation between Right and Left Eyes

The correlation between the measurements in the right and the left eyes was assessed for all the parameters by means of Pearson’s correlation factor (50 pairs of measurements). 

Influence of the Return Light Intensity

In LDF, the main expression of the intensity of the returning light is the mean DC level. The main expression of the part of the signal containing both the information pertaining to blood flow and noise is the root mean square (RMS) of the voltages, or the alternating current (AC) component. Blood flow parameters are calculated after computing a Doppler-shift power spectrum of the returning signal. The LDF parameters volume and flux, which are derived from the area under the curve of the Doppler-shift power spectrum, are standardized with DC ² (DC squared). Velocity is a ratio of flux and volume. The algorithm of the instrument requires DC levels to range between 0.5 and 5 V, which is achieved through amplification of the photocurrent. This amplification can take the following discrete values (gain): 1, 2, 5, 10, 20, and 50. Because the necessity for amplification may be related to scattering properties of the sampled tissue, and because the same tissue properties may also alter the computed LDF parameters, a new parameter, yield = DC/gain, was defined, and the relationship between yield and the LDF parameters velocity, volume, and flux were analyzed. The influence of yield on the recorded LDF parameters was partialized out in a regression model (third-order polynomial equation) applied on the logarithmic values of yield and LDF parameters. Reliability, sensitivity, and sample size calculations were repeated after partializing out the influence of yield. 

The correlation between yield in the right and the left eyes was assessed by means of Pearson’s correlation factor between all measurements in the right and left eyes (50 pairs of measurements). The reliability (R, intraclass correlation coefficient) for yield in a given eye as well as in an individual (average yield for both eyes) was calculated. 

Influence of Different Light-Scattering Properties

The influence of different light-scattering properties on the returning signal was assessed in an emmetropic model eye. Although LDF recordings in such a model eye give rise to a Doppler-shift power spectrum similar to those obtained from the choroid, the AC component of the signal consists only of noise, because there is no flow or motion. In this model eye, materials with various light-scattering properties (mono- and multilayers of semitransparent plastic folio and dyed cloths) were placed at the fundus level, and recordings were obtained without any movement and without any change in the recording settings. A range of return light intensities comparable to those observed in humans was produced by merely changing the scattering material at the fundus level. After dividing recorded RMS with gain, the obtained variable was plotted against yield, providing a graphical representation of the noise level at various yield levels. A similar plot (RMS/gain versus yield) was obtained for human data and the two plots were fitted by a least-squares regression line, depicting graphically the relative contribution of noise in the measurements obtained in human eyes. 

Influence of Correction on Blood Flow Measurement

To test the influence of the correction on the results obtained during blood flow perturbation, the choroidal LDF parameter flux was recorded under the following conditions: in two subjects, at baseline, and after various steps of increased IOP induced by means of a Langham suction cup (10 mm in diameter and a volume of 0.3 mL), and in two subjects, while they breathed a mixture of room air and 5% CO₂ through a partly closed mask system covering both mouth and nose connected to a CO₂ monitor (Capnomac Ultima; AVL Medical Systems AG, Schaffhausen, Switzerland) for approximately 15 minutes. In the latter situation, end-tidal CO₂ concentrations were monitored continuously during room air breathing and during exposure to increased inspiratory CO₂ concentrations. The relative change in blood flow was calculated with the unaltered and the corrected values obtained after applying the procedure just described. 

Results

Average (± SD) IOP was 13.6 ± 1.8 and 13.9 ± 1.2 mm Hg in right and left eyes, respectively. Systolic and diastolic blood pressure were 123.3 ± 11.8 and 77.4 ± 8.7 mm Hg, respectively. 

Although choroidal LDF measurements were always performed on the same spot (foveola), the comparison of returning light’s intensity and LDF flux values disclosed a marked variability in DC and flux values, as well as an inverse relationship between these two parameters (Fig. 1) . 

Yield and the LDF parameters velocity, volume, and flux correlated significantly in a regression model (R = 0.71, P < 0.0001; R = 0.65, P < 0.0001; and R = 0.70, P < 0.0001; for velocity, volume, and flux, respectively; Fig. 2 ). The change in yield and in LDF parameters expressed as a percentage of the average of the measurements obtained in a given eye during the 5 days demonstrated a parallel variation in yield and the LDF parameters volume and flux, whereas the LDF parameter velocity remained essentially constant over the large range of yield (Fig. 3) . 

Yields between fellow eyes correlated linearly (R = 0.53, P < 0.0001). The reliability for yield was 87% in a given eye and 78% in a given subject. Although some individuals showed yields throughout nearly the entire range in both eyes, some individuals showed clusters of yield levels (Fig. 4) . 

After the influence of yield was partialized out, reproducibility, reliability, coefficients of variation, sensitivity, and the statistical power of the technique improved markedly for the LDF parameters volume and flux, but the LDF parameter velocity parameter was affected very little (Table 1) . It must be noted, however, that the presented values refer to day-to-day variability and not to short-term variability, which would be obviously smaller with this technique. The operating characteristic curves (minimum detectable change plotted against power for an α error ≤ 0.05) improved significantly (area under the curve: P < 0.0001) after correcting for the influence of yield (Fig. 5) . 

Measurements in the model eye provided a range for yield values comparable to that obtained in human eyes. Least-squares regression lines between the AC component of the signal (after dividing it with gain) and yield demonstrated a linear relationship between these two factors in the model eye as well as in the human eyes (Fig. 6) . The two least-squares regression lines showed a clear convergence. 

Gradual increase in IOP by means of a suction cup induced a gradual decrease in choroidal LDF parameter flux in both subjects (Fig. 7) . In response to the breathing of 5% CO₂, causing an increase in end-tidal partial pressure of CO₂ from 5.2 to 7.3 kPa in one subject and from 5.3 to 6.7 kPa in the other subject, choroidal LDF parameter flux increased by 40.8% and 6.5%, respectively, as estimated with the unaltered values provided by the original algorithm of the compact, confocal choroidal laser Doppler flowmeter. The same alteration amounted to 32.8% and 15.3%, respectively, after correcting for the influence of yield. 

Discussion

The reliability of a compact choroidal laser Doppler flowmeter was investigated in healthy subjects. Choroidal blood flow measurements of the submacular choroid were obtained in both eyes on 5 days, and the relationship between the intensity of the returning light and the LDF blood flow parameters was analyzed. A new parameter called yield (yield = DC/gain) was defined. After the influence of yield was partialed out, reproducibility, reliability, coefficient of variation, sensitivity as well as statistical power of the technique improved markedly, without partializing out the information pertaining to blood flow changes. The results indicate that a major source of variability in LDF measurements is the influence of optical scattering properties in an individual’s eyes. Measurements obtained in fellow eyes suggested that the influence of yield was comparable in both eyes. Comparing results obtained from the 10 subjects with measurements obtained in a model eye demonstrated that, with decreasing yield, not only did the value of the LDF parameters increase, but the relative contribution of noise to the returning signal increased as well. In addition, the parallel variation in yield and LDF parameters demonstrated the influence of recording settings, independently of scattering properties within the eye. Recording settings influenced mostly volume and flux, but did not affect velocity. Partializing out such an influence markedly improved the performance of LDF measurements, leading to a reproducibility similar to that reported for LDF recordings of optic nerve head blood flow, ²³ although the measurements included in the present analysis had been obtained on different days. 

The indirect mode of measurement used in the present study as well as the confocal optical arrangement in the present device favor choroidal blood flow measurements. ¹⁵ According to Riva et al., ¹⁴ even in the direct mode of measurement, the contribution of the retinal capillaries during measurements within the foveola is approximately one eighth of the signal. With the indirect mode, the Doppler-shift power spectra have an exponential shape and a relatively low mean pulsatility, corresponding to a capillary vascular bed, and, thus, lending support to the assumption that the signal obtained with this method originates mostly from the choriocapillaris with little, if any, contribution form the deep feeder vessels of the choroid. ¹⁴  

The two main sources of variability of returning light intensity seem to be light-scattering properties of the sampled tissue volume and the alignment between the instrument and the eye. In the present study, each eye had systematically a different yield range compared with other eyes, reliability coefficient for yield for eyes was 87%, but this was also the case in subjects, indicated by a reliability coefficient of 78% and interocular yield correlation. These findings suggest that yield is in part defined by tissue optical properties, expectedly similar in eyes of the same person. Such an interpretation is also strengthened by the fact that the instrument used in the present study comes close to the assumption that the site of LDF measurement (foveola) does not change between recordings. 

The exact mechanism of the observed systematic dependence of LDF parameters on the returning light intensity is not clear. The major issues demonstrated in the present experiments were the alteration of the signal within eyes in the low-yield range and the dramatic increase of the relative noise contribution in the low-signal range. A possible explanation of how parameters normalized with DC (volume and flux) might be affected could be the presence of specular reflection, as observed in some other LDF applications. ²⁴ Because specular light is not shifted in frequency, it will contribute only to the DC component. However, an optical confocal arrangement, an indirect mode of measurement, and the use of polarizing filters make specular light as the major source of the observed phenomenon a remote possibility. Another speculation may be an inadequate existing DC ² -normalization algorithm for volume and flux, and some exponential factor other than 2 might be more appropriate. Explanation of the underlying physical phenomena necessitates further investigations. Empiric correction removed a large part of LDF parameter variability not directly related to blood flow, and such a correction did not partialize the information pertaining to changes in blood flow, as demonstrated in measurements during hypercapnia after increasing IOP by means of a suction cup. 

The experiments with blood flow challenge warrant some additional explanation. An increase in IOP of 12 mm Hg reduced choroidal blood flow by 30% to 40% in one of the subjects, which compares well with the range of choroidal blood flow responses found in a study demonstrating blood flow autoregulation in the human choroid. ²⁵ However, the even larger decrease in choroidal blood flow with higher IOP suggests that this subject may have an altered choroidal blood flow autoregulation. A detailed discussion of this issue would be beyond the scope of the present study. It should, nevertheless, be noted that all subjects whose choroidal blood flow was challenged in this study were completely healthy; but, obviously, autoregulatory capacity for choroidal blood flow was not comparable among them. To the best of our knowledge, altered autoregulation alone does not necessarily represent a noxious state. Many young subjects have altered autoregulation in the cerebral, ²⁶ the retinal, ²⁷ or the choroidal (Hasler et al., manuscript submitted) circulation without being in poor health. Prospective studies should evaluate the impact of altered autoregulation on long-term health. When the data obtained during hypercapnia are compared with baseline measurements, the results after correcting for yield show a difference of 8% to 9% compared with the result before correction. During continuous measurements, a change of less than 8% from baseline should not be considered relevant. ¹⁴ This limit should certainly be even higher for repeated measurements. Consequently, the choroidal blood flow change estimates before and after correcting for yield can be considered comparable. Furthermore, an average increase of 1.5% in choroidal LDF flux per 1 mm Hg increase in partial pressure of CO₂ has been described in healthy subjects. ¹⁹ The expected increase in LDF parameter flux in our two subjects amounted to 23.6% and 15.8%. Because of the large confidence interval suggested by the figures provided in the latter study (approximately ±15%), the responses found in our two subjects, especially after correcting for yield, match well the expected change during hypercapnia. 

Interindividual comparisons of baseline values are generally not recommended in LDF techniques, primarily because of varying tissue-scattering properties. If, however, these properties could find their quantitative expression in the returning light intensity, as suggested by the present data, perhaps it would be feasible to universally correct for their influence based on a sufficiently large choroidal blood flow LDF recording database and thus to enable and enhance interindividual comparisons. However, even the results of a conservative study design, which would include comparisons of baseline and follow-up measurements in the same eye, may be jeopardized by the influence of instrument alignment, particularly if baseline and follow-up measurements are performed in separate sessions. In the present study, although care was taken to reproduce accurately the recording conditions in each session, empiric correction was still warranted, suggesting that similar corrections should be run regularly. Finally, although the instrument used in this study has some special characteristics different from other continuous LDF instruments—confocal arrangement, indirect mode of measurement—it seems plausible to admit that the issues raised here may be at least partly valid for other widely used continuous LDF instrument applications and possible also scanning LDF instruments, such as the Heidelberg Retina Flowmeter (Heidelberg Engineering, Heidelberg, Germany) ²⁸ —a conjecture that should, however, be verified. 

 Supported by the Swiss National Funds for Biological and Medical Research (32-059094/99), the Fonds zur Förderung von Lehre und Forschung der Freiwilligen Akademischen Gesellschaft Basel, and the Freiwillige Akademische Gesellschaft.

 Submitted for publication March 21, 2001; revised September 4, 2001; accepted November 1, 2001.

 Commercial relationships policy: N.

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked“ advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Measurements (flux and DC) of two eyes across 5 days, during which all the recordings were obtained with the same gain setting (gain, 5). The mirror effect among the recordings between these two parameters is obvious.

Figure 2.

 

Least-squares fit defining the relationship of the LDF parameters flux (top), volume (middle), and velocity (bottom), with yield (yield = DC/gain). Each point in this scatterplot represents the mean value of 1 day’s recording in one eye (total: 20 eyes, 5 days).

Figure 3.

 

LDF parameters flux (top), volume (middle), and velocity (bottom), expressed as a percentage of the average of the measurements obtained during the 5 days, plotted against yield as a percentage of the average of the measurements obtained during the 5 days. The scatterplot demonstrates the parallel variation in yield and LDF parameters with repeated measurements, independently of scattering properties within the eye.

Figure 4.

 

Correlation of yields between fellow eyes (R = 0.53, P < 0.0001). Although some individuals showed yields throughout nearly the entire range in both eyes, suggesting that part of the variability in yield is due to recording settings on different days, some individuals seemed to form clusters of yield values, suggesting the presence of specific individual scattering properties. Different symbols indicate the 10 study participants.

Table 1.

 

Statistical Parameters before and after Correction for the Influence of Yield

Statistical Parameter	Before Correction	After Correction
Reliability (%)
Velocity	93.0	93.4
Volume	74.6	94.0
Flux	77.5	86.0
Coefficient of variation±SD (%)
Velocity	12.7 ± 14.2	11.3 ± 6.4
Volume	45.3 ± 21.5	24.5 ± 11.9
Flux	47.3 ± 23.7	25.5 ± 9.7
Minimum detectable change with 10 individuals (%)
Velocity	7.6–23.7	7.5–17.5
Volume	36.4–62.3	14.5–37.8
Flux	33.6–51.5	19.2–31.6
Sample size calculations for n individuals
Change of 5% in flux	694–1632	227–616
Change of 10% in flux	173–408	59–154
Change of 15% in flux	79–181	27–70
Change of 25% in flux	30–67	11–27
Pearson’s correlation coefficients (R) for interocular correlation (p)
Velocity	0.41 (0.003)	0.23 (0.12)
Volume	0.50 (<0.0001)	0.72 (<0.0001)
Flux	0.34 (0.014)	0.63 (0.0001)

Figure 5.

 

The operating characteristic curves (minimum detectable change plotted against power for an α error ≤ 0.05) improved significantly (area under the curve: P < 0.0001) after correcting for the influence of yield. Solid line: statistical power with native flux values; dotted line: statistical power after correcting for the influence of yield.

Figure 6.

 

The AC component of the signal (after dividing it with gain) and yield are related linearly to one another. The convergence of the least-squares fit lines obtained from the model eye, describing a signal composed of noise, and the line fitting the data obtained from the 10 subjects suggests how, with decreasing yield, the contribution of noise in the AC component of the returning signal increases.

Figure 7.

 

Choroidal LDF parameter flux decreased with increasing suction pressure applied to the eye. The result was not altered by the correction for the influence of yield.