SEARCH

SEARCH BY CITATION

Keywords:

  • fundus camera;
  • reliability;
  • repeatability;
  • reproducibility;
  • Retinal Vessel Analyser;
  • retinal vessel diameter

Abstract.

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Purpose:  Retinal Vessel Analyser (RVA) is a validated instrument to measure retinal vessel diameter in humans. The purpose of this study was to assess the reproducibility (inter-observer reliability) and the repeatability (test–retest reliability) of RVA with a microscope-mounted fundus camera to determine retinal vessel diameter in minipigs.

Methods:  Ocular fundus image from five anaesthetized minipigs was recorded in a digital videotape for approximately 5 min, under stable systemic arterial pressure and gas conditions. To evaluate the reproducibility, each one of two investigators used RVA to measure the diameter of the superior temporal retinal artery on five separate 30-second video sequences from each minipig, which were the same video sequences for both investigators. To evaluate the repeatability, one investigator performed five measurements on a single, randomly selected, 30-second video sequence from each minipig. The reproducibility was determined using the intra-class correlation coefficient (ICC), and the repeatability was assessed using the coefficient of variation (COV). Bland–Altman plots were also used to assess agreement between the two investigators.

Results:  Retinal arteriolar diameter measurements with RVA in minipigs were highly reproducible. Differences between the two investigators were lower than 0.7%. The ICC was 1.00, indicating perfect reproducibility, and the mean COV was 0.18%, reflecting excellent repeatability of the measurements with RVA.

Conclusion:  Retinal vessel diameter can reliably be determined not only in humans, but also in minipigs, using the commercially available RVA apparatus and a microscope-mounted fundus camera.


Introduction

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The knowledge of retinal haemodynamics and blood flow regulation is of fundamental importance for the understanding of mechanisms underlying retinal diseases. In ischaemic microangiopathies of the inner retina, such as diabetic retinopathy and retinal vein occlusions, impairment of retinal circulation results in blood flow alterations that, in turn, affect the delivery of oxygen and metabolic substrates necessary for the maintenance of retinal structure and function (Pournaras et al. 2008). To cope with this, various possible therapeutic agents are under investigation using protocols in animal models and in humans. Thus, there is considerable interest in studying the parameters of retinal blood flow regulation in health and in disease.

Vessel diameter constitutes an important parameter of blood flow (Pournaras et al. 2008). In the last decades, the development of non-invasive techniques to evaluate retinal haemodynamics has led to new and important information about retinal blood flow regulation in the healthy and diseased human eye. However, most of these techniques are limited by the fact that only information about blood velocity is available but not blood flow per se. Hence, it is difficult to decide whether an increase in blood velocity is caused by an increase in blood flow or by vasoconstriction within the measured vascular bed (Schmetterer & Garhofer 2007). Therefore, to determine blood flow, the exact determination of vessels diameter is crucial.

Retinal vessels diameter has been measured in past years mainly from magnified ocular fundus photographs using a calliper or by scanning across the vessel (Pournaras et al. 2008). Nowadays, Retinal Vessel Analyser (RVA), a newly developed device, has allowed online and continuous, non-invasive measurements of retinal vessels diameter (Blum et al. 1999; Lanzl et al. 2000; Nagel et al. 2000; Kiss et al. 2002). This device consists of a fundus camera, a charge-coupled device (CCD) measuring camera for electronic online image acquisition, and a computer for system control analysis and archiving of the diameter data (Garhofer et al. 2010). In general, a vessel section of about 1.5 mm in length is scanned at a frequency of 25/second (Pournaras et al. 2008).

Retinal Vessel Analyser is a validated instrument to measure retinal vessel diameter in humans (Polak et al. 2000; Pache et al. 2002; Seifert & Vilser 2002). However, a great number of protocols studying the pathophysiology of retinal circulation and testing possible therapeutic agents require the use of animal models, in which the reproducibility of RVA has not been evaluated. Therefore, the purpose of this study was to assess the reproducibility as well as the repeatability of RVA with a microscope-mounted fundus camera to determine retinal vessel diameter in an animal model.

Materials and Methods

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Animal preparation

Experiments were conducted in one eye of five minipigs (Arare Animal Facility, Geneva, Switzerland) weighing 10–12 kg. The experimental procedure was performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research.

Minipigs were prepared for the experiments as previously described (Pournaras et al. 1989; Mendrinos et al. 2008). In brief, after premedication with intramuscular injection of 3 ml (15 mg) of the tranquilliser midazolam maleate (Dormicum; Roche Pharma, Reinach, Switzerland), 3 ml (120 mg) of the tranquilliser azaperone (Stresnil; Janssen Pharmaceutica, Beerse, Belgium) and 1 ml (0.5 mg) of atropine, anaesthesia was induced with 3 mg of ketamine hydrochloride (Ketalar; Parke-Davis, Zurich, Switzerland) injected into an ear vein. Analgesia was induced with 2 ml (0.1 mg) of fentanyl (Sintenyl; Sintetica SA, Mendrisio, Switzerland), and curarization was achieved with 2 ml (4 mg) of pancuronium bromide (Pavulon; Organon SA, Pfäffikon, Switzerland). The animals were intubated and artificially ventilated. After arterial, venous and bladder catheterization, anaesthesia, analgesia and myorelaxation were maintained throughout the experiment by continuous perfusion of ketamine, fentanyl and pancuronium, respectively.

Each animal was ventilated at approximately 18 strokes/min, with a continuous flow of 20% O2 and 80% N2O, using a variable-volume respirator (Sulla 909 V; Dräger, Lübeck, Germany). Systolic and diastolic arterial blood pressure was monitored through a femoral artery with a transducer (Minograph; Siemens-Elema, Solna, Sweden). Temperature was maintained between 36 and 37 °C with a thermoblanket. Arterial partial pressure of oxygen (PaO2), arterial partial pressure of carbon dioxide (PaCO2) and arterial pH were measured from the same artery with a blood gas analyser (Labor-Systeme; Flükiger AG, Menziken, Switzerland) and kept under control by adjusting ventilatory rate, stroke volume and composition of the inhaled gas.

A head-holder was used to avoid movements because of respiration. Upper and lower eyelids were removed, as well as a rectangular area of skin surrounding the eye; the bulbar conjunctiva was detached; the sclera was carefully cleaned to 5 mm from the limbus; the superficial scleral vessels were thermo-cauterized; the globe was fixed with a metal ring sutured around the limbus; and a sclerotomy 2–3 mm posterior to the limbus was performed. A small contact lens with a flat exterior surface was placed on the cornea. The pupil was dilated with 1% atropine eye drops, and the fundus was observed using an operating microscope (Carl Zeiss Surgical GmbH, Oberkochen, Germany).

Experimental protocol

Ocular fundus image from five minipigs was filmed for approximately 5 min by a fundus camera (FF 450 Plus Fundus Camera; Carl Zeiss Meditec Inc., Oberkochen, Germany) mounted on the operating microscope, under stable systemic blood pressure, pH, PaCO2 and PaO2. Before each filming, the animals were left to stabilize for approximately half an hour, in order to achieve stable haemodynamic conditions. Repeated measurements of systemic blood pressure, pH, PaCO2 and PaO2 were performed to determine the appropriate moment for filming. A uniformly illuminated fundus image without unwanted reflections was obtained. Each filming was performed in a stable position in which the superior and inferior temporal retinal arteries, at a distance of one to two disc diameters from the optic nerve head, were clearly visible. The filmed images were stored in a digital videotape using a digital video recorder (Panasonic NV-DV 10000; Panasonic Corporation, Osaka, Japan), to be used later for the diameter measurements.

The reproducibility (or inter-observer reliability) of a device or clinical method refers to the variability in repeated measurements when one or more factors, such as observer, instrument, calibration, environment or time, are varied (McAlinden et al. 2011). In the present study, the observer was varied. The smaller this variability, the better the reproducibility is. To evaluate the reproducibility of RVA, each one of two investigators (D.N.P. and G.M.) used RVA (Imedos GmbH, Jena, Germany) to measure the diameter of the superior temporal retinal artery on five separate 30-second video sequences from each one of the five minipigs studied (i.e. a total of 25 video sequences studied by each investigator). The two investigators performed measurements on the same 30-second video sequences: these consisted of the first 30 seconds of each one of the 5 min that had previously been recorded. During each measurement, the cursor of the RVA was placed at a stable position on the vessel, so that the same length for measurement was selected by both investigators (10 cm on the screen). Each measured segment was approximately 225 pixels in length. The best imaged part of the study vessel against the background of the retina was chosen.

It has to be noted that, because the two investigators performed measurements on the same video sequences, the reproducibility (or inter-observer reliability) of the RVA system used on a specific video sequence, and not that of the whole recording process, was evaluated.

The repeatability (or test–retest reliability) of a device or clinical method refers to the variability in repeated measurements by one observer when all other factors are assumed constant (McAlinden et al. 2011). The smaller this variability, the better the repeatability is. To evaluate the repeatability of RVA, one investigator (G.M.) performed five measurements on a single, randomly selected, 30-second video sequence from each minipig.

Retinal vessel diameter determination

Retinal Vessel Analyser enables a fast, non-invasive, objective evaluation of changes in retinal vessel diameter. It consists of a fundus camera, a video camera, a real-time monitor and a personal computer with vessel diameter-analysing software for accurate vessel diameter determination. Retinal vessel diameters are analysed in real time with a maximum frequency of 50 Hz. This means that every second, a maximum of 25 readings of vessel diameter can be obtained. For this purpose, the fundus is imaged onto the CCD chip of the video camera. The observed fundus images are digitalized using a frame grabber. They can either be inspected on the real-time monitor or be stored using a videotape recorder (Garhofer et al. 2010). The latter was performed in the present study.

Because of the absorbing properties of haemoglobin, each blood vessel has a specific transmittance profile. Measurement of vessel diameter is based on adaptive algorithms using these specific profiles. To select a region of interest, each investigator defined a rectangle on the screen of the monitor. In our study, this rectangle had a length of 10 cm. Diameter was then calculated along the study vessel segment lying within the rectangle. The software calculates vessel diameter in arbitrary units (AU), which approximately correspond to micrometres (μm) at the retinal plane.

After each single measurement, the investigator had to take the mouse away from the image and begin a complete new measurement as if examining a different vessel, on a video sequence 30 second later. Transfer of the data to the Excel program of the RVA was the final step.

Statistical analysis

All data regarding vessel diameter, systemic blood pressure, pH, PaCO2 and PaO2 were expressed as the mean ± standard deviation. Statistical analysis was performed by using a commercially available statistical software package (SPSS Statistics 19; SPSS Inc., Chicago, IL, USA).

The reproducibility of RVA was evaluated using the intra-class correlation coefficient (κ), which was applied on the 25 pairs of measurements taken by the two investigators. The higher the κ value, the better the reproducibility of the method. A κ of 1 reflects perfect reproducibility, and a κ > 0.8 represents almost perfect reproducibility (Garcia-Martin et al. 2011). Bland–Altman plot analysis was also used, so as to assess the agreement between the two investigators. Of note, although Bland–Altman plotting is normally used to compare two different methods, we considered its application useful in the case of comparing the two investigators of the present study as well.

To evaluate the repeatability of RVA, the coefficient of variation was used, defined as the ratio of the standard deviation to the mean. A separate coefficient of variation was calculated for each minipig, as well as the mean of them. The lower the coefficient of variation, the better the repeatability of the method. A coefficient of variation below 5% indicates a very high repeatability (Garcia-Martin et al. 2011).

Results

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Systemic blood pressure, pH, PaCO2 and PaO2 remained practically stable during each 5-min recording from any of the animals studied, as shown in Table 1.

Table 1.   Systemic blood pressure (BP), pH, arterial partial pressure of carbon dioxide (PaCO2) and arterial partial pressure of oxygen (PaO2) at the beginning and at the end of each 5-min fundus recording from the five minipigs studied. Data regarding BP, PaCO2 and PaO2 are expressed in mmHg.
Animal0 min5 min
BPpHPaCO2PaO2BPpHPaCO2PaO2
193/787.353511595/817.3536110
2125/957.4042105118/917.4139106
3105/757.4336114102/717.4337111
4113/747.3541100111/737.3443111
584/467.43369683/457.413894

The data used to evaluate the reproducibility of the measurements with RVA are presented in Table 2. As five different retinal vessels were studied, vessel diameter values ranged from 149.1 to 214.0 AU, but those corresponding to the same video sequences were very close to each other when the two investigators were compared. The intra-class correlation coefficient (κ) of the 25 pairs of these measurements was 1.00, indicating perfect reproducibility of the method in minipigs.

Table 2.   Retinal vessel diameter in AU measured on video sequences by Investigators A and B, independently, with RVA. The intra-class correlation coefficient calculated from the 25 pairs of the corresponding measurements was 1.00, indicating high reproducibility of RVA in minipigs.
InvestigatorAnimalTime periods (min)
0–0.51–1.52–2.53–3.54–4.5
  1. AU, arbitrary units; RVA, Retinal Vessel Analyser.

A1212.9211.7212.9213.1213.5
2212.8212.6212.5214.0213.5
3149.8150.5149.5149.5149.3
4150.8150.3150.2149.6150.0
5188.3188.8187.9188.7188.2
B1212.8211.3212.5212.5213.1
2213.4212.4213.5213.9213.7
3150.2150.1150.3149.1149.1
4150.7150.6149.6150.8150.0
5189.5188.8188.8188.4188.7

Bland–Altman plot analysis showed an excellent inter-observer agreement. Differences between the two investigators were lower than 0.7%. The mean paired difference between the measurements of the two investigators was −0.1 AU, that is, near zero, and all the differences of the 25 pairs of individual measurements were within the limits of agreement (Fig. 1).

image

Figure 1.  Bland–Altman plots of the inter-observer agreement in retinal vessel diameter measurements with the Retinal Vessel Analyser. The difference ‘retinal vessel diameter by Investigator A minus retinal vessel diameter by Investigator B’ was plotted against the mean retinal vessel diameter of both investigators. The mean paired difference is shown by the solid line, the limits of agreement (±1.96 SD) by the dashed lines. An excellent inter-observer agreement is shown: the mean paired difference is near zero (i.e. −0.1 arbitrary units), and all the differences of the 25 pairs of individual measurements are within the limits of agreement.

Download figure to PowerPoint

The data used to evaluate the repeatability of the measurements with RVA are presented in Table 3. As five different retinal vessels were studied, vessel diameter values ranged from 190.0 to 230.6 AU, but those corresponding to the same animal were very close to each other. Individual coefficients of variation ranged from 0.12% to 0.24% (mean coefficient of variation = 0.18%), reflecting excellent repeatability of the measurements with RVA in minipigs.

Table 3.   Retinal vessel diameter in AU, repeatedly measured on video sequences by Investigator B with RVA. The mean coefficient of variation of the measurements was 0.18%, reflecting excellent repeatability of RVA in minipigs.
AnimalRepetitionsCOV (%)
12345
  1. AU, arbitrary units; COV, coefficient of variation; RVA, Retinal Vessel Analyser.

1190.3190.5189.3190.0190.10.24
2190.1190.8190.5190.5190.70.12
3192.1192.4192.2192.1193.10.21
4229.5230.6230.5230.1229.70.21
5190.8190.3190.3190.7190.50.14

Discussion

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The current study evaluated the reproducibility and repeatability of RVA to measure retinal vessel diameter in minipigs. The results showed perfect reproducibility and excellent repeatability of RVA when used in minipigs, indicating for the first time that retinal vessel diameter can reliably be determined in this animal model using the commercially available RVA apparatus and a microscope-mounted fundus camera. This device may thus be considered validated to be used in protocols investigating retinal vessel diameter changes under various conditions in minipigs.

The reproducibility of RVA to measure retinal vessel diameter in humans has previously been defined as high. Polak et al. (2000) performed measurements in nine healthy young volunteers and reported intra-class correlation coefficients of 0.98 and 0.96 for retinal veins and retinal arteries, respectively, concluding that short-term measurements of retinal vessel diameter with the Zeiss RVA are highly reproducible. They also calculated short-term coefficients of variation to be 1.3% and 2.6% for retinal veins and retinal arteries, respectively (Polak et al. 2000). In another study, Seifert and Vilser used RVA to measure retinal vein diameters in 12 healthy volunteers continuously for 5 min, and the measurements were repeated at the same vessel locations after 2 hr. This resulted in a short-term coefficient of variation of 1.5% (Seifert & Vilser 2002). Finally, to evaluate the short-term reproducibility of RVA, Pache et al. (2002) measured retinal vessel diameter in 20 healthy subjects at baseline and after 2 hr. Short-term variability of vessel diameter was only 1%, and the intra-class correlation coefficient was 0.97 and 0.96 for retinal veins and retinal arteries, respectively.

Thanks to the above favourable results, RVA is considered a reliable device for both analysis and follow-up of retinal vessel diameter changes in humans. The current study demonstrates that RVA can also be considered reliable for the study of retinal vessel diameter changes in minipigs. The intra-class correlation coefficient for retinal arteries was 1.00, quite comparable with that calculated in humans. Our study provides also information about repeatability, calculating a 0.18% mean coefficient of variation of repeated measurements by the same person, a number which adds to the excellent reliability of RVA in minipigs.

Some limitations about the present study need to be mentioned. Both investigators performed measurements on the same video sequences, so what was evaluated was the reproducibility (or inter-observer reliability) of RVA measurements on a specific video sequence and not that of the whole recording process (e.g. manual positioning of the animal, set-up of the microscope, etc.). In addition, retinal vessel diameter measurements were limited to retinal arterioles. This was done because, from the physiopathological point of view, changes in retinal blood flow are mainly related to changes of the retinal arteriolar diameter (Pournaras et al. 2008).

Clinical interest of the results of the current study is high. Retinal vessel response to drugs already used in ophthalmology has recently begun to draw interest (Papadopoulou et al. 2009; Lanzl et al. 2011). However, several protocols investigating physiopathological pathways of the retina, or new drugs, cannot be directly applied to humans, because of their interventional nature and safety concerns. For instance, RVA has been used in minipigs to evaluate the vasomotor effect of intra-vitreal juxta-arteriolar injection of endothelin-A receptor inhibitor (Stangos et al. 2010), l-arginine (Mendrinos et al. 2010) and l-lactate (Mendrinos et al. 2011), on retinal arteriolar diameter after acute branch retinal vein occlusion. A scientific basis for validation of RVA measurements in animals is provided by the current study.

In conclusion, retinal vessel diameter can reliably be determined not only in humans, but also in minipigs, using the commercially available RVA apparatus and a microscope-mounted fundus camera. An excellent reproducibility and repeatability of this method make it appropriate to evaluate retinal vessel diameter changes in short-term experimental protocols.

Acknowledgments

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Supported by Grant 3200B0-105809 (CJP) from the Swiss National Science Foundation. Presented at the European Association for Vision and Eye Research (EVER) 2010 Congress, Crete, Greece, October 6–9, 2010. The authors thank Nicole Gilodi for technical assistance with the conduct of this research.

References

  1. Top of page
  2. Abstract.
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References