February 2011
Volume 52, Issue 2
Free
Retina  |   February 2011
Computer-Assisted Methods to Evaluate Retinal Vascular Caliber: What Are They Measuring?
Author Affiliations & Notes
  • Helena M. Pakter
    From the Postgraduate Studies Program in Epidemiology and
    the Hospital Nossa Senhora da Conceição, Grupo Hospitalar Conceição, Porto Alegre, RS, Brazil;
  • Sandra C. Fuchs
    From the Postgraduate Studies Program in Epidemiology and
    the Divisions of Cardiology and
    the National Institute for Science and Technology for Health Technology Assessment (IATS/CNPq), Porto Alegre, RS, Brazil.
  • Marcelo K. Maestri
    From the Postgraduate Studies Program in Epidemiology and
    Ophthalmology, Hospital de Clinicas de Porto Alegre, Porto Alegre, RS, Brazil; and
  • Leila B. Moreira
    From the Postgraduate Studies Program in Epidemiology and
    the Divisions of Cardiology and
    the National Institute for Science and Technology for Health Technology Assessment (IATS/CNPq), Porto Alegre, RS, Brazil.
  • Luciana M. Dei Ricardi
    Ophthalmology, Hospital de Clinicas de Porto Alegre, Porto Alegre, RS, Brazil; and
  • Vítor F. Pamplona
    the Informatics Institute, UFRGS (Universidade Federal do Rio Grande do Sul), Porto Alegre, RS (Rio Grande do Sul), Brazil;
  • Manuel M. Oliveira
    the Informatics Institute, UFRGS (Universidade Federal do Rio Grande do Sul), Porto Alegre, RS (Rio Grande do Sul), Brazil;
  • Flávio D. Fuchs
    the Divisions of Cardiology and
    the National Institute for Science and Technology for Health Technology Assessment (IATS/CNPq), Porto Alegre, RS, Brazil.
  • Corresponding author: Sandra C. Fuchs, Centro de Pesquisa Experimental, 2° andar, CARDIOLAB, Instituto de Avaliação de Tecnologias em Saúde (IATS), Hospital de Clínicas de Porto Alegre, Universidade Federal do Rio Grande do Sul, Ramiro Barcellos, 2350, 2° andar, 90.035-003, Porto Alegre, RS, Brazil; scfuchs@terra.com.br
Investigative Ophthalmology & Visual Science February 2011, Vol.52, 810-815. doi:10.1167/iovs.10-5876
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to Subscribers Only
      Sign In or Create an Account ×
    • Get Citation

      Helena M. Pakter, Sandra C. Fuchs, Marcelo K. Maestri, Leila B. Moreira, Luciana M. Dei Ricardi, Vítor F. Pamplona, Manuel M. Oliveira, Flávio D. Fuchs; Computer-Assisted Methods to Evaluate Retinal Vascular Caliber: What Are They Measuring?. Invest. Ophthalmol. Vis. Sci. 2011;52(2):810-815. doi: 10.1167/iovs.10-5876.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose.: Computer-assisted methods to measure retinal vessel diameters have been incorporated into research, but it is not clear which component of the vessels they are measuring. This study was conducted to compare measurements of retinal vessel diameter by using imaging-processing software on color fundus photographs (FPs) and fluorescein angiographs (FAs).

Methods.: FP and FA images were taken simultaneously in 52 eyes of 31 patients referred for angiography for diagnosis of retinal disease. Arteriolar and venular calibers were measured in two concentric zones around the optic disc center. Pearson correlation coefficients and Bland-Altman plots were used to evaluate the agreement between the measurements made by FP and FA.

Results.: The differences between the diameters measured by the microdensitometric method from FP and FA were 2.59 ± 8.67 μm in the inner arteriola, 4.93 ± 7.47 μm in the outer arteriola, −1,58 ± 8.49 μm in the inner venula, and −1.80 ± 7.28 μm in the outer venula. The differences plotted by the Bland-Altman method were slight. The Pearson correlation coefficients of measurements by FP and FA were 0.84 for inner zone and 0.87 for outer zone arterioles and 0.93 and 0.94 for the inner and outer zone venules, respectively.

Conclusions.: The very slight differences between measurements of retinal vessel diameter by the two methods demonstrate that the microdensitometric method mostly measures the vessel lumen. Differences in vessel diameters measured by the microdensitometric method observed in clinical conditions may therefore be ascribed to variation in wall thickness or vasoconstriction.

Digital processing of images obtained by retinography techniques developed in recent years 1,2 has improved the measurement of retinal vessel diameters and its application in the investigation of the mechanisms and stratification of the risk for cardiovascular diseases. Retinal arteriolar and venular diameters and their relationship have been associated with the incidence of hypertension 3 and severe hypertension, 4 heart failure, 5 coronary artery disease, 6,7 stroke, 8 and cardiovascular mortality. 9 These observations have been consistently reported since computer-assisted methods were introduced for retinal vessel measurements, with great reproducibility 1,9 in digitized fundus phonographs from large cohort population studies. 2,10 12  
While the computer-assisted methods, mainly the microdensitometric ones, have improved retinal vascular diameter assessment in color digital fundus photography (FP), it has not yet been determined whether these methods actually measure the outer vessel diameter (vessel wall and lumen), central red blood cell column, or even the overall lumen. The premises are that, as in other vessels, blood cells remain in the central part of the retinal vessel, whereas the outer column, with a slower stream, is essentially made of plasma. Color FP seems to show the blood column of the retinal vessels, whereas the slower peripheral plasma stream and the vessel walls are transparent to direct visualization. 13  
Fluorescein angiography (FA) is the technique of choice for visualizing a blood column, but it is an invasive, laborious, and potentially harmful examination, and thus it cannot be used to assess vascular calibers in large subject samples and could be replaced by indirect methods if the precise correlation between them could be established. The comparison between vascular caliber measured by image-processing software in color FP and FA has not been properly assessed. An assessment of simultaneous measurements of retinal vessel diameter by both methods would allow estimating whether the indirect method measures the central blood flow, the lumen, or the entire vessel. The purpose of this study was to determine which component or components of vessels are measured in images taken in microdensitometric retinography, taking FA as the gold standard. 
Methods
Sample
Patients referred for FA to an ophthalmology clinic in Porto Alegre, southern Brazil, were enrolled to participate in this cross-sectional study. Participants had indication for FA for clinical conditions such as retinal disease. The tenets of the Declaration of Helsinki were followed, and informed consents were signed by all participants. The study was approved by the Ethics Committee of Hospital de Clínicas of Porto Alegre, which is accredited by the Office of Human Research Protection as an Institutional Review Board. 
Image Acquisition
FPs were captured with a digital fundus camera system (model TRC 50 EX and MT-10 Cameras; Topcon Co. Japan; and Angioimage, ver. 1.51; Retinal Imaging System, Buenos Aires, Argentina), at a 35° angle. Twenty minutes after pharmacologic pupil dilation with tropicamide 1% drops, color FPs were obtained, including optic disc–centered images. After this, 5 mL of 10% fluorescein sterile dye (Ophthalmos Farm Ind., São Paulo, Brazil) was injected intravenously into the antecubital vein, and FA was performed in the usual way, according to the clinical indication. 
For this study, we obtained an optic disc–centered image during the FA procedure. The image was captured at the moment that both arterioles and venules were full of fluorescein (arterial–venous phase and were storage in bitmap file format, with an original size of 5.62 Mb each. 
Alignment of Image Pairs Organization
The selected optic disc–centered acquired images were organized in pairs (color FP-FA) and cropped to a common resolution (1.616 × 1.216 pixels). For alignment, the operator marked at least three pairs of corresponding points on each image pair (i.e., at least three points on the retinography, and their counterparts on its associated angiography). Such correspondence was established in the valleys of vessel bifurcations, defining a transformation that aligned the vessels of an image pair. The transformation matrix was computed solving a system of linear equations obtained from the image coordinates of the corresponding points. With the alignment of retinography–angiography pairs, we were able to compute the average vessel caliber for any given vessel segment in retinography and angiography at the same time. 
Retinal Vessel Measurement
At the opening of an image pair in the software, four concentric rings were automatically dropped over it. The inner circle measured the equivalent of 1800 μm (using the calibration factor of 7.73 μm/pixel [px], described below), and it was centered at the optic disc head. The center of the optic disc was computed as the average coordinates of the set of pixels in the brightest regions (intensity bigger than 250) in the green channel of the image. The diameter of the optic disc in each image was computed as the largest continuous bright pixel segments along horizontal and vertical pixel scans. The second, third, and fourth rings have 1.5, 2, and 3 disc diameters, respectively. If, in this case, the program aligned the rings incorrectly at the optic disc, the operator could adjust them (Fig. 1). After adjustment of the images for the magnification differences by using a calibration factor, two concentric zones are delimited: the inner zone (A), ranging from one half of the disc diameter to one full disc diameter; and the outer zone (B), between one and two disc diameters from the margin. Arteriolar and venular calibers were measured in these two areas around the optic disc center. Details of the method have been published elsewhere. 1  
Figure 1.
 
The four steps of our software: (i) user selects a ROI, (ii) software finds the vessel, (iii) software calculates the medial axis, and (iv) the final vessel border with calibers calculated at every axial point. A, inner zone; B, outer zone.
Figure 1.
 
The four steps of our software: (i) user selects a ROI, (ii) software finds the vessel, (iii) software calculates the medial axis, and (iv) the final vessel border with calibers calculated at every axial point. A, inner zone; B, outer zone.
In measuring the vessel caliber, the updated version of the software comprises four steps: (1) definition of the region of interest (ROI), (2) calculation of the vessel boundaries, (3) determination of vessel orientation, and (4) vessel diameter measurement (Fig. 2). 
Figure 2.
 
Software window showing a venula (v10) being measured in the inner zone of an FP (top) and at an FA (bottom). For each pair of FP and FA, a maximum of 24 arteriole and 24 venule segments were identified and measured (12 of each in the inner zone, and 12 of each in the outer zone). The measurements results are presented at the left. Highlighted here, inside the yellow box, venular diameter as measured in the FP (117.14 ± 5.59 μm/ maximum deviation of 13.81 μm; 94 measurements done in this segment) and as measured in FA (115.60 ± 4.61 μm/ maximum deviation of 13.10 μm; 94 measurements done in this segment). For each segment measured, the statistics were sent to a database automatically.
Figure 2.
 
Software window showing a venula (v10) being measured in the inner zone of an FP (top) and at an FA (bottom). For each pair of FP and FA, a maximum of 24 arteriole and 24 venule segments were identified and measured (12 of each in the inner zone, and 12 of each in the outer zone). The measurements results are presented at the left. Highlighted here, inside the yellow box, venular diameter as measured in the FP (117.14 ± 5.59 μm/ maximum deviation of 13.81 μm; 94 measurements done in this segment) and as measured in FA (115.60 ± 4.61 μm/ maximum deviation of 13.10 μm; 94 measurements done in this segment). For each segment measured, the statistics were sent to a database automatically.
Once the user provided a rectangular ROI, the software computed the vessel boundaries, using the green channel of the retinography. The vessel boundaries were found by using the wide-line detector model of Liu et al. 14 This model reduces the width of the smooth transitions (blur) between the background color and the vessel color and then sets the vessel edge as the spatial middle point in this transition space. The reduction of smooth regions allowed greater reliability and reproducibility of the results. Noise and defocus frequently found in low-quality retinography did not significantly affect the measurements. Our implementation uses the bright- and dark-line configurations for angiograph and retinograph images, respectively. As in the detector model of Liu et al., the radius of the circular mask was set to 25% of the optical disc radius, and the brightness contrast threshold was set to be the standard deviation of the luminance values in the ROI. After the detection of the vessels, the software applied a noise-removal filter using mathematical morphology to remove thin lines and some artifacts generated by high-frequency retinal background textures. 
Since vessels are not straight lines, the measurements must take into account the changes in vessel orientation in the images. Therefore, the software extracted the medial axis for each vessel. The medial axis was the set of pixels inside the vessel that are equidistant from the vessel borders. It can be seen as the skeleton of the vessel. Each axial pixel was the source of a measurement, which was performed perpendicular to the orientation of the medial axis at that point. Thus, the number of measurements made in an ROI is directly proportional to the image resolution. 
Determination of the Scale Factor
To increase the precision of the absolute values of vessel caliber measurements, we used a calibration factor to adjust for the magnification differences due to camera optics, image resolution, and the patient's refractive errors. The calibration factor is the ratio of micrometers per pixel in a definite length of a digitized image of the eye fundus. It was introduced as a calculation constant in the microdensitometric software before the measurements of the vascular calibers were obtained. 
The calibration factor was determined by measuring the distance between the center of the macula to the center of the optic disc in a sample of 20 color photographs containing the macula and the optic disc in the same field, using a customized caliper tool in our software. By assuming that the average disc diameter is 1800 μm 15 and the distance from the center of the optic disc to the center of the macula is 2.5 disc diameters, 16 or 4500 μm, we calculated the scale factor by the following formula:   For this study, the calibration factor was 7.73 μm/px. 
Reliability of the Method
The reproducibility of the method was tested by an ophthalmologist who assessed the same photographs twice and by an independent ophthalmologist who verified a subsample of 20% of the photos, to determine intraobserver and interobserver reliability, respectively. The reliability of the method was determined by using the intraclass correlation coefficient (ICC). 
Statistical Analysis
The caliber of the vessels was defined by the mean value of each vessel segment measured in the color retinography and its paired FA. We excluded from the analysis the vessel segments that had only measurements from one type of image. These limitations were due to artifacts or blurriness of one of the images. Pearson correlation coefficients and Bland-Altman plots were used to evaluate the agreement between the measurements done by FP and FA. In the Bland-Altman approach, the differences between the two techniques (FP and FA) were plotted against the averages of the two techniques. Horizontal lines were drawn at the mean difference, and at the limits of agreement, which were defined as the mean difference ± 1.96 times the standard deviation of the differences. 17  
Results
A total of 46 patients (83 eyes) were examined: 16 volunteers (31 eyes) were excluded, and 52 eyes of 30 participants were analyzed. The main exclusion criteria were media opacities, diabetic retinopathy grade 2 or worse, venous or arterial occlusions, or the technical limitations described earlier (Table 1). 
Table 1.
 
Characteristics of the Study Population and the Clinical Diagnosis
Table 1.
 
Characteristics of the Study Population and the Clinical Diagnosis
Age, y (mean ± SD) 57.2 ± 8.7
Male sex, n (%) 14 (47)
White skin color, n (%) 29 (97)
Diagnosis, n eyes (%)
    Normal 30 (58)
    Diabetic retinopathy, grade 1 12 (23)
    Glaucoma 3 (6)
    Peripapilar atrophy 2 (4)
    Serpiginous choroidopathy 2 (4)
    Choroidal nevus 1 (2)
    Chronic central serous retinopathy 1 (2)
    Cicatricial toxoplasmosis retinochoroiditis 1 (2)
Mean arteriolar and venular calibers in inner and outer zones, in micrometers, as measured in color FPs and FAs are shown in Table 2. The inner zone mean arteriolar caliber of 94.34 ± 15.1 μm and mean venular caliber of 110.2 ± 18.7 μm observed in the color FPs were virtually the same as those of the measurements of the FAs (mean arteriolar lumen of 91.8 ± 15.7 μm and mean venular lumen of 111.6 ± 22.3 μm). The same occurred in the outer zone. 
Table 2.
 
Mean Arteriolar and Venular Calibers in the Inner and Outer Zones, as Measured by FP and FA
Table 2.
 
Mean Arteriolar and Venular Calibers in the Inner and Outer Zones, as Measured by FP and FA
n Mean SD Range Minimum Maximum
Inner arteriola
    FP 124 94.34 15.10 67.64 59.98 127.62
    FA 124 91.75 15.69 83.51 59.28 142.79
    ΔFP-FA 124 2.59 8.67 66.78 −36.22 30.56
Outer arteriola
    FP 127 94.04 14.71 77.19 56.79 133.98
    FA 127 89.11 14.70 84.94 54.58 139.52
    ΔFP-FA outer 127 4.93 7.47 37.98 −14.79 23.19
Inner venula
    FP 147 110.02 18.68 89.08 61.59 150.67
    FA 147 111.60 22.30 96.69 58.61 155.30
    ΔFP-FA inner 147 −1.58 8.49 43.86 −20.09 23.77
Outer venula
    FP 159 107.66 18.42 91.02 54.88 145.90
    FA 159 109.46 21.39 96.89 57.93 154.82
    ΔFP-FA outer 159 −1.80 7.28 46.74 −20.43 26.31
The Pearson correlation coefficients of measurements done by FP and FA were 0.84 for the inner zone and 0.87 for the outer zone for arterioles, and 0.93 and 0.94 for venules, respectively. Intra- and interobserver agreement was excellent, with ICCs ranging from 0.94 to 0.99 (Table 3). Bland-Altman plots and regression lines between the measurements in FP and for FA of the inner and outer arterioles and venules are presented in Figures 3 to 6. The differences in measurements by both methods were slight. Arterioles tended to have larger diameters when measured by FP and venules when measured by FA. 
Table 3.
 
ICC with 95% CI for Intra- and Interobserver Reliability
Table 3.
 
ICC with 95% CI for Intra- and Interobserver Reliability
ICC (95% CI) Interobserver ICC (95% CI) Intraobserver
FA inner arteriola 0.975 (0.953–0.986) 0.978 (0.978–0.993)
FP inner arteriola 0.977 (0.962–0.986) 0.950 (0.902–0.975)
FA inner venula 0.988 (0.979–0.993) 0.998 (0.996–0.999)
FP inner venula 0.978 (0.962–0.987) 0.983 (0.968–0.991)
FA outer arteriola 0.957 (0.930–0.973) 0.974 (0.958–0.984)
FP outer arteriola 0.941 (0.901–0.964) 0.941 (0.900–0.964)
FA outer venula 0.986 (0.978–0.991) 0.991 (0.982–0.995)
FP outer venula 0.991 (0.986–0.995) 0.978 (0.947–0.989)
Figure 3.
 
Bland-Altman plot for inner arterioles. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus average and its 95% confidence interval (CI) is also shown (y = 6.44 + 0.041x, P = 0.19).
Figure 3.
 
Bland-Altman plot for inner arterioles. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus average and its 95% confidence interval (CI) is also shown (y = 6.44 + 0.041x, P = 0.19).
Figure 4.
 
Bland-Altman plot for the outer arterioles. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus the average and its 95% CI is also shown (y = 4.89 + 0.0004x; P = 0.26).
Figure 4.
 
Bland-Altman plot for the outer arterioles. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus the average and its 95% CI is also shown (y = 4.89 + 0.0004x; P = 0.26).
Figure 5.
 
Bland-Altman plot for the inner venules. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus the average and its 95% CI is also shown (y = 18.70 + −0.18x; P = < 0.0001).
Figure 5.
 
Bland-Altman plot for the inner venules. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus the average and its 95% CI is also shown (y = 18.70 + −0.18x; P = < 0.0001).
Figure 6.
 
Bland-Altman plot for the outer venules. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus the average and its 95% CI is also shown (y = 14.86 + −0.15x, P = < 0.0001).
Figure 6.
 
Bland-Altman plot for the outer venules. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus the average and its 95% CI is also shown (y = 14.86 + −0.15x, P = < 0.0001).
Discussion
This study showed that retinal vessel diameters measured in color FPs were quite similar to those measured in FA images. These results were observed both in arterioles and venules. To the best of our knowledge, these findings concerning the use of a computer-assisted method of retinal vessel measurement have not been reported. 
In paraffin-embedded sections of human retina, the major arteries have a luminal diameter of ∼100 μm near the optic disc, and a wall thickness of ∼18 μm at this point. 18 Even though our measurements were made at least 1 disc diameter from the optic disc margin, where the vessels are naturally narrower than at the margin, they could be considered representative of in vivo vessel diameters. As the differences of arterioles caliber measurements between FP and FA images were less than 5 μm (<3 μm for arterioles and <5 μm for venules, even in larger arterioles), it is highly probable that vessel walls are not measured by FP microdensitometric methods. Recently, Michelson et al. 19 have shown a noninvasive method to estimate the wall thickness of small retinal vessels by scanning laser Doppler flowmetry. With this technique, they found a mean retinal temporal superior artery diameter of 110 ± 16.8 μm and mean retinal temporal superior vein diameter of 134 ± 20.1 μm. Flow diameters were 82.3 ± 13.4 μm for the artery and 97.9 ± 18.3 μm for the vein. By this, they estimated the thickness of each side of the artery walls to be 14.0 ± 5.3 μm and the thickness of each side of the vein walls to be 18.1 ± 7.5 μm. Although our measurements in FPs of the inner area (closer to the optic disc) were very similar to those in their study, they were wider for the lumen, which were acquired from a conventional digital fundus camera different from the scanning laser doppler flowmeter. 
Since the measures in FA images included the complete blood column (i.e., central red blood cells and peripheral plasma stream), a wider vessel diameter would be expected in these images than in color FP images, which are mainly composed of the central red blood stream. In this study, the calibers were virtually the same, since we found very slight differences in the arteriolar and venular diameters for in both FP and FA images. This fact could point out that the retinal vessel diameters measured in color FPs are representative of the complete vessel lumen as in FAs. A possible explanation for not finding wider vessel caliber in FA images is a consequence of Poiseuille's Law, as Guyton states: “in small vessels essentially all the blood is near the vessels walls. So, the extremely rapidly flowing central stream of blood almost does not exist.” 20 As retinal vessels are small, the central red blood cells and the peripheral plasma stream would be closer together, filling the vessel lumen with a more homogeneous blood column. In fact, in retinal vessels, red blood cells show higher velocity at the center of the vessel, in a parabolic velocity profile, and thus the blood is more homogeneously concentrated across the vessel. 21 The impossibility of measuring the retinal vessel wall in color fundus photographs could also be related to the fact that the wall is transparent to the light, whereas the blood reflects it, and therefore color images tend to miss the walls. In addition, since the vessel has a cylindrical shape, the light that shines at the edge of the vessel was partially scattered, decreasing the possibility of detecting the vessel wall exactly. 
Although the difference from the measurements in FP and FA were minimal, it is of interest to note that they work in distinct ways. While the arterioles have larger diameters in the measurements of FPs, the venules have larger diameters in the measurements of FAs. Also, by analyzing the regression line shown with the Bland-Altman plots, we noticed that in the case of venules, as the average of the measurements increased, the deltas (Δ) were even more negative. This result is probably due to image artifacts, such as differences in color contrast or brightness that could have interfered with the software readings. Since there are differences between arteriolar and venular color and brightness, the venular blood column appears as a darker red in FP, and the software could more easily detect the boundary between the blood column and the background, while in the arterioles, it has a brighter red color, with the border becoming slightly blurred against the background, and thus the vessel caliber could be slightly overestimated. Another point could be the effect of the histologic differences of the arteriolar versus venular walls on the light scattering, as mentioned above, resulting in a wider arteriolar caliber. 
The influence of the variation due to the cardiac cycle was a matter of concern. 22 24 The pulse wave difference between the two images (FP and FA) could bias the measurements. Since our software does multiple measurements along a vessel segment, potential influences of the cardiac cycle were lessened. In addition, the small size of retinal vessels (∼50–200 μm) could be a reason that measurements are not completely accurate. 13 We used images larger than 5 Mb that had satisfactory resolution of the vessel measurements, and we excluded vessels with less than 50 μm. Other possible explanations of the moderate and nonsignificant differences that we observed are the degree of systemic autonomic nerve stimulation 25 and the degree of fundus pigmentation (Hubbard LD, et al. IOVS 1992;33:ARVO Abstract 804). One could consider some measuring bias of the retinopathies that were indicated in the FAs, but we emphasize that the objective was a vessel-to-vessel comparison between the fundus images pairs. 
In reviewing recently published articles, we found only one study that compares retinal vessel diameters in paired color and FA and FPs. 26 The authors acknowledge that the accuracy of their approach has not been properly validated. While their technique seems to reasonably recover the structure of the retinal vessel network, it computes only the average diameter of any vessel segment, which should not include any branching structures. Our approach, on the other hand, computes the diameters of a vessel segment at as many cross sections as there are pixels in its skeleton (several hundred in a typical segment). As a result, we can produce more reliable statistics about the vessel diameters and, consequently, more reliable comparisons between the pairs of vessels imaged by angiography and retinography techniques. 
Taking the disc diameter of 1.8 mm as a basis for the calibration could raise concerns about the measure's precision, as it has associations with ethnic background 27 and could be influenced by the equipment (Elledge JA, et al. IOVS 2005;46:ARVO E-Abstract 2583). As in this study 97% of the participants were white, we assumed the disc diameter to be 1.8 mm because it is the medium value observed for white people in one study, 27 being close to the best estimated value for in vivo measurements (Elledge JA, et al. IOVS 2005;46:ARVO E-Abstract 2583). We did not compare the diameter measured manually in older studies with the gold standard. Since the microdensitometric method was validated against manual measurement, 1 it is to be expected that the correlation would be similar to that reported in this study. We compared the diameter of vessels without focal narrowing. Since there is concordance in this context, it is expected that the parallelism persists within narrowed segments, but this could be checked only in newer software. 
In conclusion, our computer-assisted method of retinal vessel analysis by a microdensitometric method of color retinography was able to precisely measure the complete blood column that fills the vessel lumen, including the central red blood cells and the peripheral plasma stream. Studies of correlations between vessel diameters and cardiovascular outcomes should therefore assume that microdensitometric methods of color FPs measure only the complete vessel lumen. 
Footnotes
 Supported, in part, by National Counsel of Technological and Scientific Development (CNPq) Grant 485831/2007-4 and Hospital de Clínicas de Porto Alegre-FIPE (Fundo de Incentivo a Pesquisa e Esino) Protocol GPPG 04-465, RS, Brazil
Footnotes
 Disclosure: H.M. Pakter, None; S.C. Fuchs, None; M.K. Maestri, None; L.B. Moreira, None; L.M. Dei Ricardi, None; V.F. Pamplona, None; M.M. Oliveira, None; F.D. Fuchs, None
References
Pakter HM Ferlin E Fuchs SC . Measuring arteriolar-to-venous ratio in retinal photography of patients with hypertension: development and application of a new semi-automated method. Am J Hypertens. 2005;18:417–421.
Sharrett AR Hubbard LD Cooper LS . Retinal arteriolar diameters and elevated blood pressure: The Atherosclerosis Risk in Communities Study. Am J Epidemiol. 1999;150:263–270.
Wong TY Shankar A Klein R Klein BEK Hubbard LD . Prospective cohort study of retinal vessel diameters and risk of hypertension. BMJ. 2004;329:79.
Smith W Wang JJ Wong TY . Retinal arteriolar narrowing is associated with 5-year incident severe hypertension: The Blue Mountains Eye Study. Hypertension. 2004;44:442–447.
Wong TY Rosamond W Chang PP . Retinopathy and risk of congestive heart failure. JAMA. 2005;293:63–69.
Duncan BB Wong TY Tyroler HA Davis CE Fuchs FD . Hypertensive retinopathy and incident coronary heart disease in high risk men. Br J Ophthalmol. 2002;86:1002–1006.
Wong TY Klein R Sharrett AR . Retinal arteriolar narrowing and risk of coronary heart disease in men and women: The Atherosclerosis Risk in Communities Study. JAMA. 2002;287:1153–1159.
Wong TY Sharrett AR Couper DJ . Retinal microvascular abnormalities and incident strokes. The Atherosclerosis Risk in Communities Study. Lancet. 2001;358:1134–1140.
Wong TY Klein R Nieto FJ . Retinal microvascular abnormalities and 10-year cardiovascular mortality: a population-based case-control study. Ophthalmology. 2003;110:933–940.
Ikram MK Witteman JCM Vingerling JR Breteler MMB Hofman A de Jong PTVM . Retinal vessel diameters and risk of hypertension: The Rotterdam Study. Hypertension. 2006;47:189–194.
Wang JJ Liew G Klein R . Retinal vessel diameter and cardiovascular mortality: pooled data analysis from two older populations. Eur Heart J. 2007;28:1984–1992.
Wong TY Islam FMA Klein R . Retinal vascular caliber, cardiovascular risk factors, and inflammation: The Multi-Ethnic Study of Atherosclerosis (MESA). Invest Ophthalmol Vis Sci. 2006;47:2341–2350.
Patton N Aslamc T MacGillivrayd T . Retinal image analysis: concepts, applications and potential. Prog Retin Eye Res. 2006;25:99–127.
Liu L Zhang D You J . Detecting wide lines using isotropic nonlinear filtering. Image Process IEEE Trans. 2007;16:1584–1595.
Jonas JB Gusek GC Naumann GO . Optic disc, cup and neuroretinal rim size, configuration and correlations in normal eyes. Invest Ophthalmol Vis Sci. 1988;29:1151–1158.
Synthanayothin C Boyce JF Cook HL Williamson TH . Automated localisation of the optics disc, fovea and retinal blood vessels from digital colour fundus images. Br J Ophthalmol. 1999;83:902–910.
Bland JM Altman DG . Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1(8476):307–310.
Hogan MJ Alavarado JA Weddel JE . Histology of the Human Eye: an Atlas and Textbook. In: Philadelphia: WB Saunders Co.; 1971:508–521.
Michelson G Warntges S Baleanu D . Morphometric age-related evaluation of small retinal vessels by scanning laser Doppler flowmetry: determination of a vessel wall index. Retina. 2007;27:490–498.
Guyton AC Hall JE . Overview of the circulation: medical physics of pressure, flow, and resistance. In: Textbook of Medical Physiology. 11th ed. Philadelphia: Elsevier Saunders; 2006:161–170.
Logean E Schmetterer L Riva C . Velocity profile of red blood cells in human retinal vessels using confocal scanning laser Doppler velocimetry. Laser Phys. 2003;13:45–51.
Chen HC Patel V Wiek J Rassam SM Kohner EM . Vessel diameter changes during the cardiac cycle. Eye. 1994;8:97–103.
Dumskyj MJ Ishii N Nishihara Y Horie A . The accurate assessment of changes in retinal vessel diameter using multiple frame electrocardiograph synchronised fundus photography. Curr Eye Res. 1996;15:625–632.
Knudtson MD Klein BEK Klein R . Variation associated with measurement of retinal vessel diameters at different points in the pulse cycle. Br J Ophthalmol. 2004;88:57–61.
Baer R Hill D . Retinal vessel responses to passive tilting. Eye. 1990:751–756.
Martinez-Perez M Hughes AD Thom SA Bharath AA Parker KH . Segmentation of blood vessels from red-free and fluorescein retinal images. Med Image Anal. 2007;11:46–61.
Varma R Tielsch JM Quigley HA . Race-, age-, gender-, and refractive error–related differences in the normal optic disc. Arch Ophthalmol. 1994;112:1068–1076.
Figure 1.
 
The four steps of our software: (i) user selects a ROI, (ii) software finds the vessel, (iii) software calculates the medial axis, and (iv) the final vessel border with calibers calculated at every axial point. A, inner zone; B, outer zone.
Figure 1.
 
The four steps of our software: (i) user selects a ROI, (ii) software finds the vessel, (iii) software calculates the medial axis, and (iv) the final vessel border with calibers calculated at every axial point. A, inner zone; B, outer zone.
Figure 2.
 
Software window showing a venula (v10) being measured in the inner zone of an FP (top) and at an FA (bottom). For each pair of FP and FA, a maximum of 24 arteriole and 24 venule segments were identified and measured (12 of each in the inner zone, and 12 of each in the outer zone). The measurements results are presented at the left. Highlighted here, inside the yellow box, venular diameter as measured in the FP (117.14 ± 5.59 μm/ maximum deviation of 13.81 μm; 94 measurements done in this segment) and as measured in FA (115.60 ± 4.61 μm/ maximum deviation of 13.10 μm; 94 measurements done in this segment). For each segment measured, the statistics were sent to a database automatically.
Figure 2.
 
Software window showing a venula (v10) being measured in the inner zone of an FP (top) and at an FA (bottom). For each pair of FP and FA, a maximum of 24 arteriole and 24 venule segments were identified and measured (12 of each in the inner zone, and 12 of each in the outer zone). The measurements results are presented at the left. Highlighted here, inside the yellow box, venular diameter as measured in the FP (117.14 ± 5.59 μm/ maximum deviation of 13.81 μm; 94 measurements done in this segment) and as measured in FA (115.60 ± 4.61 μm/ maximum deviation of 13.10 μm; 94 measurements done in this segment). For each segment measured, the statistics were sent to a database automatically.
Figure 3.
 
Bland-Altman plot for inner arterioles. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus average and its 95% confidence interval (CI) is also shown (y = 6.44 + 0.041x, P = 0.19).
Figure 3.
 
Bland-Altman plot for inner arterioles. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus average and its 95% confidence interval (CI) is also shown (y = 6.44 + 0.041x, P = 0.19).
Figure 4.
 
Bland-Altman plot for the outer arterioles. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus the average and its 95% CI is also shown (y = 4.89 + 0.0004x; P = 0.26).
Figure 4.
 
Bland-Altman plot for the outer arterioles. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus the average and its 95% CI is also shown (y = 4.89 + 0.0004x; P = 0.26).
Figure 5.
 
Bland-Altman plot for the inner venules. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus the average and its 95% CI is also shown (y = 18.70 + −0.18x; P = < 0.0001).
Figure 5.
 
Bland-Altman plot for the inner venules. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus the average and its 95% CI is also shown (y = 18.70 + −0.18x; P = < 0.0001).
Figure 6.
 
Bland-Altman plot for the outer venules. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus the average and its 95% CI is also shown (y = 14.86 + −0.15x, P = < 0.0001).
Figure 6.
 
Bland-Altman plot for the outer venules. The differences between the measurements acquired in two types of images (FP − FA) are plotted against the averages of the measurements. The regression line of difference versus the average and its 95% CI is also shown (y = 14.86 + −0.15x, P = < 0.0001).
Table 1.
 
Characteristics of the Study Population and the Clinical Diagnosis
Table 1.
 
Characteristics of the Study Population and the Clinical Diagnosis
Age, y (mean ± SD) 57.2 ± 8.7
Male sex, n (%) 14 (47)
White skin color, n (%) 29 (97)
Diagnosis, n eyes (%)
    Normal 30 (58)
    Diabetic retinopathy, grade 1 12 (23)
    Glaucoma 3 (6)
    Peripapilar atrophy 2 (4)
    Serpiginous choroidopathy 2 (4)
    Choroidal nevus 1 (2)
    Chronic central serous retinopathy 1 (2)
    Cicatricial toxoplasmosis retinochoroiditis 1 (2)
Table 2.
 
Mean Arteriolar and Venular Calibers in the Inner and Outer Zones, as Measured by FP and FA
Table 2.
 
Mean Arteriolar and Venular Calibers in the Inner and Outer Zones, as Measured by FP and FA
n Mean SD Range Minimum Maximum
Inner arteriola
    FP 124 94.34 15.10 67.64 59.98 127.62
    FA 124 91.75 15.69 83.51 59.28 142.79
    ΔFP-FA 124 2.59 8.67 66.78 −36.22 30.56
Outer arteriola
    FP 127 94.04 14.71 77.19 56.79 133.98
    FA 127 89.11 14.70 84.94 54.58 139.52
    ΔFP-FA outer 127 4.93 7.47 37.98 −14.79 23.19
Inner venula
    FP 147 110.02 18.68 89.08 61.59 150.67
    FA 147 111.60 22.30 96.69 58.61 155.30
    ΔFP-FA inner 147 −1.58 8.49 43.86 −20.09 23.77
Outer venula
    FP 159 107.66 18.42 91.02 54.88 145.90
    FA 159 109.46 21.39 96.89 57.93 154.82
    ΔFP-FA outer 159 −1.80 7.28 46.74 −20.43 26.31
Table 3.
 
ICC with 95% CI for Intra- and Interobserver Reliability
Table 3.
 
ICC with 95% CI for Intra- and Interobserver Reliability
ICC (95% CI) Interobserver ICC (95% CI) Intraobserver
FA inner arteriola 0.975 (0.953–0.986) 0.978 (0.978–0.993)
FP inner arteriola 0.977 (0.962–0.986) 0.950 (0.902–0.975)
FA inner venula 0.988 (0.979–0.993) 0.998 (0.996–0.999)
FP inner venula 0.978 (0.962–0.987) 0.983 (0.968–0.991)
FA outer arteriola 0.957 (0.930–0.973) 0.974 (0.958–0.984)
FP outer arteriola 0.941 (0.901–0.964) 0.941 (0.900–0.964)
FA outer venula 0.986 (0.978–0.991) 0.991 (0.982–0.995)
FP outer venula 0.991 (0.986–0.995) 0.978 (0.947–0.989)
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×