J Hum Hypertens.2014 May;28(5):310-315.

Relation Between Blood Pressure Control and Changes in Retinal Microcirculation in A Group Of Hypertensive Patients.

Pose-Reino A1; Pena-Seijo M1, Diaz Diaz JL2; Díaz Peromingo JA3; Suarez Tembra M4; Monte Secades R5; González Vázquez S6; González Penedo M6; Méndez-Naya I7; Estévez Nuñez JC7; Coll-de-Tuero G8; Gomez-Ulla Irazazábal F9.


Author Affiliations:

  1. Department of Internal Medicine, Hospital de Conxo.Medicine Deparment, Complexo Hospitalario Universitario de Santiago de Compostela.Spain.
  2. Department of Internal Medicine. Hospital Abente Lago, ACoruña. Spain.
  3. Department of Internal Medicine. Hospital Barbanza, Ribeira.Spain.
  4. Department of Internal Medicine. Hospital San Rafael, A Coruña. Spain.
  5. Department of Internal Medicine. Hospital LucusAugusti, Lugo.Spain.
  6. Department of Electronics and Computation (Artificial Vision Group). University of A Coruña.
  7. Department of Econometrics.University of Santiago de Compostela. Spain.
  8. Assistance Healthcare Institute Anglès, Girona. Spain.
  9. Department of Ophthalmology, Hospital de Conxo.Complexo Hospitalario Universitario de Santiago de Compostela. Spain.



Aims: We report the results of hypertensive treatment over the retinal arteriole narrowing in a group of 189 hypertensive patients during a six-month treatment for hypertension, which were included in an observational study in clinical practice condition and analyzed in digital photographs of eye fundus, with previously described semiautomatic snakes methods.

Methods and results: We included 189 hypertensive patients, 74.0% of patients had BP under control after 6 months of follow-up. The arteriovenous ratio (AVR), measured as the relation between the average retinal arteriole and venous diameter, significantly increased after six months of follow-up in both eyes: right eye AVR: 0.769 + 0.065 vs. 0.799 + 0.066, (p< 0.0001), left eye AVR: 0.770 + 0.065 vs. 0.796 + 0.071 (p<0.0001). AVR changes were caused by increases in arteriole diameter. No lineal correlationwas found between blood pressure lowering levels and AVR increase.

Conclusion: We conclude that in our group of hypertensive patients it was possible to increase the retinal arteriole diameter, expressing an improvement of retinal microcirculation after six months of adequate BP control. Although there is an inverse tendency between blood pressure and arteriovenous ratio changes, we could not find a lineal correlation between theses changes.

PMID: 24257513



To our group, the study of the retina presents both scientific and clinical interest, as the retina is an easily observable part of the central nervous system.

The first alteration we observe in the fundus of the eye of hypertensive patients consists of a narrowing of the artery calibre and indeed as retinal microcirculation can be easily observed and assessed non-invasively, it offers an opportunity for detailed in vivo study of the structure of small arterioles and venules. And so,it has been described in association with processes such as hypertension and diabetes. Some vascular changes in the eye fundus, such as the arteriolar constriction or the arteriovenous nicking, are early signs of cardiovascular diseases . Studies have shown that these lesions could be associated with increased cardiovascular mortality in persons with hypertension and related to an increased number of cerebral white matter lesions and cognitive impairment. Retinal microvascular lesions are also seen in people without a history of hypertension and could predict incident stroke independently of blood pressure.

Changes in the eye fundus observed in most hypertensive patients are subjective, imprecise and poorly reproducible. For this reason, we are focus on finding a method to assess objectively these changes without interobserver variability which allows measuring vascular abnormalities. We have previously developed a computer program to process images for evaluation of retinal vascular caliber in a quantitative and reproducible way. In a subsequent study we have described a semiautomatic system model based on snakes model, with greater reproducibility and reliability.

Description of the method:

Our technique is based on the application of a software for tanalysis of a digital photograph of eye fundus centred on the papilla and with no mydriasis. This allows for a remote analysis of any photograph of a patient that is made with a non-mydriatic retinal camera under these conditions.

The vessels are segmented and measured in several circumferences concentric to the optic nerve. The resulting vessel segments at each radius are classied as artery or vein independently by an expert physician. Then, the user selects the set of vessels suitable to estimate the arteriovenous ratio (AVR) as the ratio between the average artery and vein vessel calibers.

In this study, we show the effects telmisartan (plus hydrochlorothiazide if necessary to control blood pressure) has on the retina microvasculature in 189 uncontrolled hypertensive patients over a period of six months analyzed with the semiautomatic method based on snakes.

The main objective of the study was to see the changes in retinal microcirculation measured as average arteriole diameter, average venular diameter, and average AVR, after six months of treatment using software previously validated by our team.

Results of study:

Changes in the AVR were caused by increases in artery diameter, right eye: (5.998±1.539) vs. (6.225±1.638), p<0.0001; and left eye :(6.084±1.630) vs. (6.287±1.654), p:0.0027. There is a significant increase in the venous diameter of the right eye, (7.817±1.859) vs. (7.837±2.066), p: 0.0025, with no significant changes in the left eye: (7.905±1.912) vs. (7.907±1.997).

Average baseline systolic BP was 154.6±14.6 mmHg and 133.6±12.0 mmHg at the end of the study. Average baseline diastolic blood pressure was 91.7± 9mmHg and 79.5±7.7mmHg at the end of the study.

The percentage of patients who showed an increase in retinal AVR between the final visit and the baseline visit was in 56% of patients.

Although a tendency was observed that associated the decrease in BP throughout follow-up and changes to retinal AVR, no statistically significant correlation could be found between the quantitative changes in both variables. This suggests that, as well as an adequate control of BP, there are other non established variables that may have an influence on changes to retinal microcirculation.


Figure 1. Phases of the methodology.

The importance of this study:

In this study, we confirm that it is possible to detect an increase in the retinal arteriole diameter, which suggests a modification of the structure of retinal arterioles in a short period of time i.e., six months, with a previously validated, simple, objective and economical method. Although there is an inverse tendency between blood pressure and arteriovenous ratio changes, we could not find a correlation between these changes. In other words, a greater decrease in BP throughout the study could not be associated in any statistically significant manner with a greater increase in AVR at the end of the study, although a decrease in BP seems to be a necessary condition for an increase in retinal AVR.

At the begining of the study there were no differences in blood pressure levels between the patients that had received a treatment with diuretics and those who had not, but the only factor that could be associated in the multi-variant analysis to the change in the AVR was the treatment with diuretics previous to the patients’ inclusion in the study.

There could be two reasons to explain this, diuretics and beta-blocker are known to be the anti-hypertensive drugs with the least effect on general microcirculation. These patients’ previous treatment may not have affected the wall of the vessel and the change in treatment to a renin-angiotensin system blocker could have rendered this effect more apparent. The other possible explanation would be that following the diuretic treatment the renin-angiotensin system would activate in the retinal microcirculation, which would place retinal microcirculation in optimal conditions for the effect of its blocking.


Figure 2: Expert classication of each vessel segment represented by a circle of radius equal to the vessel width measured by the application. Arteries are shown in red and veins in blue. Numbers are just vessel segment identiers.


In recent months our group is working on the classification of retinal vessels for automatic distinction between arterioles and venules.

Any process of automatic classification of a collection of elements requires two steps. First, select a criterion that makes distinguishable the different types of elements of the set and, secondly, to determine which algorithm to distinguish the elements of each type with a higher rate of success based on the criteria chosen.

Regarding the classification of retinal vessels, colour is the key feature for sorting, as arteries are clearer than veins. Thus, from the eye fundus eye image and vessel segmentation, different metrics to characterize each vessel segment based on color have been calculated using the multiscale Retinex algorithm. (Fig.1)

Algorithm selection depends on the application domain. In our case, it should be noted that images lighting is not constant and retinal fundus pigmentation varies from one individual to another. This makes necessary to use a clustering algorithm for the classification. These types of algorithms divide elements into a fixed number of groups according to differences in the characteristics of these elements. The proposed methodology uses the k-means algorithm.

This algorithm does not apply to the image as a whole, but it is divided into four quadrants centered on the optical disc, and clustering algorithm is applied separately to the vessel segments found in each quadrant. These four quadrants are rotated to cover all possible combinations, repeating the clustering process on each one. The division into quadrants and the rotation process minimize the effects of non-uniform illumination.

Every clustering process assign to a vessel segment the label of artery, vein or unclassified. Each vessel segment accumulates different labels obtained and, finally, the most frequent label is assigned.

To ensure the label of each vessel is applied as well to a tracking algorithm which allows connecting the vessel segments throughout that vessel. Thus, vessel segments are connected through the vasculature, which makes possible to ensure each vessel classification.


Figure 3: From left to right and top to bottom: manual expert classification, automatic classification algorithm using k-means on rotated quadrants, vessel segments connected by shortest path algorithm, final classification of the vessels applying the full methodology.

This methodology has been validated on a set of 100 images evaluated by two experts and another set of 58 images evaluated by three experts, considering vessel segments in which all experts agree as gold standard for validating. In the first one, the experts coincidence level in vessels classification is 97.37%, and in the second one 96.53%. In the first case, the proposed methodology has a success rate in classification of 87.68% and in the second case a success rate in classification of 93.23%. (Fig 2-3)

After validation of automatic method, we are trying to incorporate the automatic study of retinal images to clinical practice in hypertensive and diabetic patients to confirm the prognostic value.



1. Pose-Reino A, Pena-Seijo M, Diaz Diaz JL, Díaz Peromingo JA; Suarez Tembra Monte Secades R et al. J Hum Hypertens.2014 May;28(5):310-315.

2. A. Pose-Reino, M. Pena, M. G. Penedo, M. Ortega, M. Rodríguez-Blanco, P. Vega, J.L. Díaz, N. Fernández, J.C. Estévez, F. Gómez-Ulla, “Estimation of the retinal microvascular calibre in hipertensive patients with the snakes semiautomatic model. Medicina Clínica, 135(4), 145-150, 2010.

3. S.G.Vázquez, N.Barreira, M.G.Penedo, M.Ortega, A.Pose-Reino. “Improvements in Retinal Vessel Clustering Techniques:Towards the Automatic Computation of the Arterio Venous Ratio”, Computing, 90(3-4), 197-217, 2010.



This study was partly financed by Project ETES-FIS 08/90459, of the Fondo de Investigación Sanitaria (FIS). Spanish Ministry of Science & Innovation and Project PS08/05 of the Department of Health of the Xunta de Galicia.



Prof. Antonio Pose Reino.

Department of Internal Medicine.

Hospital de Conxo. Complexo Hospitalario Universitario de Santiago de Compostela.

Rúa Ramón Baltar s/n 15706. Santiago de Compostela (A Coruña). Spain

e-mail: antonioposereino@gmail.com




Multiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier SchönmannMultiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier Schönmann