REPRODUCIBILITY OF THE HEIDELBERG RETINA FLOWMETER BY AUTOMATIC FULL FIELD PERFUSION IMAGE ANALYSIS*

MICHELE IESTER,1,2 MICHELE ALTIERI,1 GEORG MICHELSON,3 PAOLO VITTONE,2 CARLO E. TRAVERSO1 and GIOVANNI CALABRIA1

1Department of Neurological Sciences, Ophthalmology, Genetic, Clinica Oculistica, University of Genoa, and 2Division of Ophthalmology, G. Gaslini Institute; Genoa, Italy; 3Department of Ophthalmalogy and Eye Hospital, Friedrich-Alexander-University Erlangen-Nürnberg, Erlangen, Germany

Abstract

Purpose: To evaluate the intra-observer reproducibility of an automatic full field perfusion image analysis (AFFPIA) for assessing retinal blood flow. Methods: One eye each of ten subjects was recruited at random. Blood flow measurements were analyzed using the Heidelberg Retina Flowmeter (HRF), which calculates the flow, volume, and velocity of each pixel. Intra-observer reproducibility was calculated for the AFFPIA program. The superior and inferior sections of the optic nerve heads (ONHs) were considered. Retinal blood flow was calculated for each of the following parts: temporal, nasal, and rim area. Results: The intra-observer intra-image AFFPIA coefficient of variation (COV) ranged from 0.5-5% in the temporal area, from 0.1-5.3% in the nasal area, and from 0.5-28% in the rim area. The intra-observer inter-image AFFPIA COV of flow measurements ranged from 1-7.3% in the temporal area, from 1.5-10% in the nasal area, and from 2-30% in the rim area. Conclusion: Retinal blood flow measured by HRF and analyzed by AFFPIA has good intra-observer reproducibility in both the temporal and nasal areas.

Introduction

Indirect clinical data suggest that vascular factors are involved in the pathophysiology of glaucomatous optic neuropathy.1-3 Several technologies compete to allow clinicians to evaluate ocular blood flow.4-6

In this study, we tested the Heidelberg Retina Flowmeter’s (HRF) reproducibility using the automatic full field perfusion image analysis (AFFPIA) program, which can evaluate the topography of perfused capillary vessels of the retina and optic nerve head.

*The details of this study have been published in the Journal of Glaucoma, 2002.

Address for correspondence: Michele Iester, MD, Viale Teano 71/1, 16147 Genoa, Italy. Email: m_iester@ hotmail.com

Perimetry Update 2002/2003, pp. 293–297

Proceedings of the XVth International Perimetric Society Meeting, Stratford-upon-Avon, England, June 26–29, 2002

edited by David B. Henson and Michael Wall

© 2004 Kugler Publications, The Hague, The Netherlands

294 M. Iester et al.

Patients and methods

Five normal subjects and five glaucomatous patients were included in the study. Patients with diabetes, systemic hypertension, or vasospastic symptoms were excluded. Patients with an area of beta atrophy were excluded in order to avoid bias due to the local hyperreflectivity of the sclera. No patients with ametropia of more than ±5 diopters, lens opacity, or other ocular disease were included in the study. The diagnosis of primary open-angle glaucoma was based on the presence of typical glaucomatous visual field defects and optic disc damage.7 In all cases, the visual field was analyzed by Humphrey Field Analyzer, program 30-2, full threshold, SITA (Humphrey Inc, Leandro, CA). Blood pressure and heart rate were also assessed.

HRF combines the principles of a confocal scanning laser and a laser Doppler flowmeter (670 nm; 100 µW). After 128 scans of each examined point, HRF calculates a bi-dimensional map of the laser Doppler shift within a 300-µm slice of tissue, over a rectangular area (5° x 20°) of the posterior pole of the eye. The calculation is performed using a fast Fourier transformation. The laser Doppler shift values recorded at different locations are displayed on a monitor in a color code image. A frequency shift is calculated for each pixel.8-12

Raw data were analyzed using the AFFPIA program. The details of this technique have been published elsewhere.13 Briefly, AFFPIA calculates the Doppler frequency shift and the hemodynamic factor or flow of each pixel, according to the theory of Bonner and Nossal.8 For the valid estimation of retinal blood flow, some assumptions must be made: adequate brightness, no ocular movement, Doppler shift of less than 2000 Hz. In order to meet these requirements, the resulting perfusion image is processed to account for underand overexposed pixels, saccades, and the retinal vessel tree. With AFFPIA, the operator marks saccades and the location of the rim area; in a further step, the capillaries and large vessels of the retina are automatically identified by a vessel detection algorithm based on the intensity and the perfusion image. Underand overexposed pixels and the saccades are automatically excluded. The local inhomogeneities of the perfusion map are softened by a moving average procedure, performed with a size of 5 x 5 pixels.13 The time needed for capturing each image is two seconds. This time allows averaging across the cardiac circle for areas larger than a pixel.

The blood flow was automatically analyzed in the temporal, nasal, and rim areas. The heart-beat associated pulsation of capillary blood flow was accounted for by plotting the mean capillary flow of each horizontal line against time.

Three HRF images of the superior area and inferior area were taken and analyzed for each patient. Attempts were made to ensure that these three images were not precisely superimposed. All the images included either the superior or inferior ONH section. When the images were analyzed, temporal and nasal peripapillary retina and optic rim area flows were calculated.

In order to determine the intra-image/intra-observer reproducibility, the first HRF perfusion map of each patient was analyzed by the same observer (MI) five times, and the coefficient of variation (COV) calculated. When the inter-image/intra-observer reproducibility was studied, the same observer (MI) analyzed three consecutive perfusion maps of the same area once.

Results

The mean age of the subjects was 68.1 ± 5.2 years (mean ± standard deviation) and the mean refractive error –0.9 ± 1.9 diopters. The mean visual field mean deviation was –3.1 ± 2.7 and the mean corrected pattern standard deviation 3.3 ± 1.6. Mean blood pressure was 146.5/80 (systolic/diastolic blood pressure), and the mean heart rate 80.7.

When the intra-image/intra-observer reproducibility of the superior sector was studied in the healthy group, COV was 3.6% (1.8; mean (standard deviation) for the temporal area, 17.1% (10.2) for the rim area, and 3.3% (1.9) for the nasal area. In the glaucomatous group, intra-image/intra-observer COV was 3.9% (1.7) for the temporal area, 17.4% (9) for the rim area, and 4.5% (0.7) for the nasal area.

When the intra-image/intra-observer reproducibility of the inferior sector was studied in the healthy group, COV was 4.2% (0.8) for the temporal area, 21.8% (4.6) for the rim area, and 3.5% (1.1) for the nasal area. In the glaucomatous group, intra- image/intra-observer COV was 4.0% (1.6) for the temporal area, 20.8% (5.9) for the rim area, and 4.6% (0.4) for the nasal area.

When the inter-image/intra-observer reproducibility of the superior sector was studied in the healthy group, COV was 5.4% (2.6) for the temporal area, 21.1% (11.1) for the rim area, and 7.0% (3.2) for the nasal area. In the glaucomatous group, inter- image/intra-observer COV was 6.6% (0.7) for the temporal area, 26.7% (4.5) for the rim area, and 8.4% (1.6) for the nasal area.

When the inter-image/intra-observer reproducibility of the inferior sector was studied in the healthy group, COV was 6.6% (0.7) for the temporal area, 18.4% (10) for the rim area, and 6.8% (3.4) for the nasal area. In the glaucomatous group, inter- image/intra-observer COV was 7.0% (0.2) for the temporal area, 20.3% (8.7) for the rim area, and 7.8% (1.7) for the nasal area.

No difference in reproducibility was found between the superior and inferior areas or between normal subjects and glaucomatous patients.

Discussion

Using HRF technology, Michelson, Schmauss et al. found the reliability coefficients of flow, volume, and velocity to be 0.84, 0.85, and 0.84, respectively, suggesting that there was only a 15-16% of variability due to the intrinsic errors component.12

Nicolela et al. found an interobserver reliability coefficient ranging from 0.90-0.98. The intersession reproducibility was good when the measurements were carried out using a 10 x 10 pixel box, while reproducibility was poorer for the retina, rim area, and lamina cribrosa when the 4 x 4 pixel box was used. The COV ranged from a median of 17-34%.11

Kagemann et al. showed that intersession reproducibility was best using pixel-by- pixel analysis of the entire image and the COV of intrasession reproducibility of repeated measures averaged 9.6% for volume, 6.5% for velocity, and 6.6% for total blood flow, while the COV of intersession reproducibility averaged 30% for all the parameters.9

Griesser et al. showed that the 50 x 50 pixels box had the higher reliability, how-

296 M. Iester et al.

ever, HRF results were influenced by height when measurements were made in the optic disc.10

Using the AFFPIA program to analyze HRF measurements, Michelson et al. noted that a coefficient of reliability of 0.74 for intra-observer reproducibility and 0.95 for inter-observer reproducibility were found.13

The reproducibility of the technique did not change significantly when the superior or inferior sector was considered. An increase in COV was found when the interimage reproducibility was tested. In the inter-image study, the three consecutive HRF images were not taken precisely in the same area, in order to challenge the reproducibility. This is reflected in the size of the peripapillary area measured, changing from 3.2 to 11.2 in the temporal and nasal areas, the rim area, defined by an outer and inner edge, changed more than the other areas. As suggested by others,11 we measured the flow of the rim area using images acquired when focusing on the superficial retina; this is because of the low reflectivity of the rim making focusing difficult. Furthermore, variability can be due to the difficulty in determining the outer and inner rim margins; our intra-image study showed that the analyzed rim area varied from 22 to 40%. These factors may explain the greater range of measurements found in the neuroretinal rim. The possible influence of a different observer was not considered in this study.

Since the reproducibility of HRF has been shown to be questionable, several authors tried to find different ways to use HRF measurements.9-13 Using AFFPIA, the peripapillary retinal flow is measured globally, independently of the cardiac cycle during the measurement. When a 10 x 10 pixels box is used to measure retinal blood flow, it is important to know in which cardiac phase the measurement is obtained, since the flow can increase or decrease accordingly. When the flow is measured using many more points, as with AFFPIA, systolic and diastolic phases are averaged, and the mean retinal flow will include the entire cardiac circle. No significant difference was found between normal and glaucomatous eyes regarding any of the reproducibility values.

Our results with AFFPIA show limitations in the interpretation of the rim values, and they were not applicable to the lamina cribrosa. In the future, it will be possible to evaluate serial measurements and the practical use of this technology.

Acknowledgment

The authors have no proprietary interest in the development or marketing of any of the products mentioned in this article.

References

1.Flammer J: The vascular concept in glaucoma. Surv Ophthalmol 38:3-6, 1994

2.Hayreh SS, Zimmerman MB, Podhajsky P, Alward WL: Nocturnal arterial hypotension and its role in optic nerve head and ocular ischemic disorders. Am J Ophthalmol 117:603-624, 1994

3.Graham SL, Drance SM, Wijsman K, Douglas GR, Mikelberg FS: Ambulatory blood pressure monitoring in glaucoma: the nocturnal dip. Ophthalmology 102:61-69, 1995

4.Riva CE, Harino S, Petrig BL, Shonat RD: Laser Doppler flowmeter in the optic nerve. Exp Eye Res 55:499-506, 1992

5.Harris A, Sergott RC, Spaeth GL, Katz JL, Shoemaker JA, Martin BJ: Color Doppler analysis of ocular vessel blood velocity in normal tension glaucoma. Am J Ophthalmol 118:642-649, 1994

6.Galassi F, Nuzzaci G, Sodi A, Casi P, Capelli S, Vielmo A: Possible correlation of ocular blood flow parameters with intraocular pressure and visual field alterations in glaucoma: a study by means of color Doppler imaging. Ophthalmologica 208:304-308, 1994

7.European Glaucoma Society: 1998 Terminology and Guidelines for Glaucoma, ch 2, pp 64-65. Savona: Dogma 1998

8.Michelson G, Schmauss B: Two dimensional mapping of the perfusion of the retina and optic nerve head. Br J Ophthalmol 79:1126-1132, 1995

9.Kagemann L, Harris A, Chung HS, Evans D, Buck S, Martin B: Heidelberg retinal flowmetry: factors affecting blood flow measurement. Br J Ophthalmol 82:131-136, 1998

10.Griesser SM, Lietz A, Orgul S, Schotzau, Hendrickson P, J Flammer, Haefliger IO: Heidelberg Retina Flowmeter parameters at the papilla in healthy subjects. Eur J Ophthalmol 9:32-36, 1999

11.Nicolela MT, Hnik P, Schulzer M, Drance SM: Reproducibility of retinal and optic nerve head blood flow measurements with scanning laser Doppler flowmetry. J Glaucoma 6:157-164, 1997

12.Michelson G, Schmauss B, Langhans MJ, Harazny J, Groh MJM: Principle, validity and reliability of scanning laser Doppler flowmetry. J Glaucoma 5:99-105, 1996

13.Michelson G, Welzenbach J, Pal I, Harazny J: Automatic full field analysis of perfusion images gained by scanning laser Doppler flowmetry. Br J Ophthalmol 82:1294-1300, 1998