European Association for the Study of Diabetic Eye Complications
Annual Meeting Munich 21-23rd May 2004
Screening & Epedemiology
SCREENING AND EPIDEMIOLOGY-IMAGING TECHNOLOGY SCREENING FOR DIABETIC RETINOPATHY: AUTOMATIC DETECTION OF MICROANEURYSMS IN COLOR FUNDUS IMAGES A. Erginay(1), P. Massin(1), T. Walter(2), J.C. Klein(2), A. Gaudric(1)
(1)Department of Ophthalmology, Lariboisière Hospital, University Paris 7, Paris, France
(2)Center of Mathematical Morphology, School of Mines, Paris, France
Purpose: To present an automated fundus photographic image-analysis algorithm for the detection of microaneurysms and the assessment of its performance. Due to a lack of specialists compared to a high and increasing number of diabetic patients, such an algorithm can be used for the computer assisted mass screening of diabetic retinopathy.
Material and Methods: The automatic algorithm works on the green channel of the color image. After a prefiltering step, dark details are extracted by means of the diameter closing and an automatic noise dependent threshold. Features, calculated for these candidates, allow their classification into real microaneurysms and false positives. 57 images were used to assess the quality of the automatic detection of microaneurysms. They were taken with a SONY color video 3CCD camera on a Topcon TRC 50 IA retinograph; resolution was 640 x 480. The images were then graded manually by two independent human graders. A golden standard was established comparing their results. The result of microaneurysms counting obtained by the automatic algorithm was then compared to this golden standard.
Results: The 57 images contained 814 microaneurysms marked manually (golden standard). 16 of the 57 images contained less than 4 microaneurysms. The comparison between the result obtained by the algorithm and the golden standard gave a mean sensitivity of 88.1% with 2.3 false positives per image.
Conclusion: In the framework of computer assisted mass-screening of diabetic retinopathy, an algorithm for automatic detection of microaneurysms in color fundus images has been presented. The results have been compared to the results obtained by two specialists with satisfying results.
DETECTION OF DIABETIC RETINOPATHY- A COMPARISON BETWEEN RED-FREE DIGITAL IMAGES AND COLOUR TRANSPARENCIES
Gunvor von Wendt 1) , Paula Summanen3) , Kerstin Hallnäs 2) , Peep Algvere1) , Kauko Heikkilä4) , Stefan Seregard1)
1, 2) St Eriks Eye Hospital, Stockholm and Örnsköldsvik Hospital, Sweden
3, 4) Helsinki University Central Hospital and Dept of Public Health, Helsinki University, Finland
Aims To compare how diabetic retinopathy (DRP) was detected from red-free (r-f) digital images versus colour transparencies
Methods Two ophthalmologists graded using an ETDRS- based grading scale, 60° r-f digital images and colour transparencies of 107 diabetic patients. The discordantly scored eyes were graded by the graders together to get a consensus level of DRP for each method. The eyes with discordant consensus results were further graded using all available photographic material to get a final consensus of DRP. Inter-method variations were presented as percentages and using kappa (k) and weighted kappa (wk) statistics. The errors of the two consensus gradings with respect to the final consensus grading were compared using Mc Nemar's test.
Results Agreement between the individual and the consensus grading results from colour transparencies occurred in 93 % (k = 0.90, wk = 0.97) and 86 % (k = 0.79, wk 0.88) and from r-f digital images in 88 % (k = 0.83, wk = 0.96) and 84 % (k = 0.78, wk 0.91) for grader 1 and grader 2, respectively. Agreement between methods was obtained in 76/107 eyes (71 %; k = 0.58 and wk = 0.79). In the 31 discordantly graded eyes the level of DRP was underestimated in 20/31 (64 %) versus 7/31 eyes (23 %) and overestimated in1/31 (3 %) versus 3/31 eyes (10 %) from colour transparencies and r-f digital images, respectively. The error tendencies were significantly lower when using r-f digital images (p < 0.008)
Conclusion R-f digital images can be recommended for screening of diabetic retinopathy.
EVALUATION OF A SCREENING PROGRAM FOR DIABETIC RETINOPATHY BY FUNDUS PHOTOGRAPHS IN PRIVATE PRACTICE
A. Erginay1 , P. Massin1 , A. BenMehidi 1 , E. Eschwege2 , J.-P. Aubert 3 , M. Marre4 , F. Fagnani 5 ,S Tcherny6 , M Bouhassira6 ;
1 Ophthalmology Department, Lariboisière Hospital, Paris, France,
2 INSERM, Paris, France
3 Réseau de Santé Paris Nord, Paris, France
4 Endocrinology Department, Bichat Hospital, Paris, France
5 CEMKA, Paris, France
6 Lilly, Suresnes, France.
Objective: The main aim of this study is to evaluate diabetic retinopathy (DR) screening by eye fundus photographs taken by a non-mydriatic retinograph compared to a dilated eye exam by an ophthalmologist.
Methods : Two networks of general practitioners (GPs) (control group= ARES 92, experimental group = Réseau de Santé Paris Nord) were recruited on voluntary basis from each network. Each group was to enroll 500 patients based on the following inclusion criterion: patients with diabetes but without documented DR or a dilated eye exam performed within the last year. The experimental group was referred to the retinograph for eye fundus photographs without pupil dilation. The control group was referred directly to an ophthalmologist for a dilated eye exam. The photographs, taken by an orthoptist, were transmitted electronically to a central analysis department and read by an ophthalmologist (Lariboisière Hospital). The main criterion of screening evaluation was the existence of an eye fundus exam report in the GP file after the 6 months of follow up.
Results: Physician participation percentage was 35,4% in the Paris Nord Network and 12,1% in the ARES 92 network. Respectively, 456 and 426 patients were included in the experimental group and the control group. 39 patients (23 of which did perform the eye fundus photography) from the PARIS-NORD group and 9 from the ARES 92 group were lost by the GPs in the follow-up period. Among the 417 patients included in the Paris Nord group, the physician did receive the report of the Topcon photography examination for 309 patients (74.1%). Among these patients, 49 (15.9%) were affected by DR, of which 15 (4.9%) were in an unrecognized severe stage of DR, while 16 (5.2%) patients had results that were non- interpretable. In the control group, the physician received the ophthalmologist report for 298 (71.5%) patients in the 6-month follow-up period. Among these patients, 31 (10.4%) were affected by DR, of which 22 (7.3%) were in a severe stage.
Conclusion: The screening of DR by a non-mydriatic camera, performed in an organized network, can be an effective means for identifying and staging DR.
DETECTING DIABETIC RETINOPATHY BY AUTOMATED IMAGE ANALYSIS
Anja Bech Hansen, Niels Væver Hartvig, Maja Skytte Jensen, Henrik Lund-Andersen, Michael Larsen.. Department of Ophthalmology, Herlev Hospital, University of Copenhagen.
Purpose: To investigate the use of automatic image analysis for detecting diabetic retinopathy in retinal images captured with a digital non-mydriatic camera.
Methods: 82 patients (163 eyes) with either type 1 or type 2 diabetes mellitus were recruited based on medical records. All patients were photographed after the installation of mydriatic eye drops using a Topcon TRC-NW6S non-mydriatic digital fundus camera interfaced with a JVC 3CCD color camera (pixel resolution of 1450 x 1026). Five overlapping non- stereoscopic 45° images were captured of each eye. Two independent readers performed a masked grading using the ETDRS scale. In cases of disagreement a third retinal specialist served as an adjudicator. The images were also analyzed using commercial fundus image-analysis software. The system uses advanced modeling of the gray-level image function of digital images, primarily the green color channel, and provide automated red microaneurysm and hemorrhage lesion detection. The automatic image analysis was performed on the five individual images of each eye, and eyes with at least one detected lesion were classified as “Possible DR” by the system.
Results: In total 42 eyes were assessed to have no diabetic retinopathy (No DR or Questionable DR) by the human graders, and 121 eyes to have diabetic retinopathy (Minimal DR or worse). A ROC curve demonstrating the range between sensitivity and specificity of the automated red-lesion- detection in detecting eyes with diabetic retinopathy was constructed: the AUC of the ROC curve was 90.8 %. Setting the visibility threshold to 2.1 for the automated algorithm, the sensitivity of detecting eyes with diabetic retinopathy was 90.1 % and the specificity of detecting eyes without diabetic retinopathy was 73.8 %.
Conclusion: Our results indicate that automated detection of red lesions in digital fundus images may be used as a first-step screening tool for classifying diabetic patients with respect to the presences or absence of diabetic retinopathy demonstrating a potential for substantially reducing the burden of manual grading.
OCCURRENCE OF DIABETIC RETINOPATHY AMONG THE PATIENTS WITH SECONDARY DIABETES AND DIABETES TYPE 1, (COMPARATIVE STUDY)
Tomas Sosna1,2 , Zuzana Vlasakova2 , Zdenek Benes 1 , Ivana Vyhnankova1
Thomayer Memorial University Hospital, Prague1, Institute for Clinical and Experimental Medicine, Prague2
Purpose: The aim of this comparative study was to assess the occurrence of diabetic retinopathy (DR) in 40 patients with secondary - pancreatogenic diabetes and 40 patients with Type 1 diabetes. We analyzed the frequency of DR in relation to metabolic compensation and other related factors.
Methods: The study comprised 80 diabetic patients treated at our departments from 1988. 40 secondary-pankreatogenic diabetics, predominantly alcoholic genesis, and 40 diabetics Type 1. Patients cohort were matched for sex, age, and diabetes onset (secondary, 31+/- 9 [+/-SD] years versus Type 1, 32 +/- 8 years) and duration of diabetes (9+/-5 versus 10 +/- 6 years). Both groups were treatded by insulin intensified regimen. DR was assessed from stereo pair photography modified EURODIAB study. Furthermore, we evaluate in both groups HbAlc level, body mass index, mirkoalbuminuria, proteinuria, urea, serum creatinine, blood pressure, blood lipids and manifestation of neuropathy. We also addressed family history of diabetes and DR.
Results: The prevalence of DR was significantly higher in Type 1 patients then those with secondary diabetes (50/20%). More than 1/3 Type l. diabetic patients had proliferative form of DR, whereas sec. diabetics have no proliferative form of DR. Need for laser surgery was merely in Type l. patients. HbAlc levels and insulin requirement had no statistical differences between both groups. Other monitored data and risk factors showed no obvious differences.
Conclusion: Data on microvascular complication in pancreatogenic DM is scarce. In contrast to the majority of reported results from other centers we found in our patients very low frequency of DR. The plausible explanation might be well timed, better and tight compensation of DM, despite the fact that alcoholic patients usually have lower compliance. However, a longer follow up of these patients is necessary.
INFLAMMATION AND VASCULAR ENDOTHELIAL DYSFUNCTION ARE ASSOCIATED WITH RETINOPATHY THE HOORN STUDY
Manon vonHecke, Jacqueline M. Dekker, Giel Nijpels, Annette C. Moll, Robert J. Heine, Lex M. Bouter, Coen D.A. Stehouwer, Bettine C.P. Polak Department of Ophthalmology & Institute for Research in Extramural Medicine, VU University Medical Center, Amsterdam, The Netherlands
Purpose The exact pathogenesis of retinopathy in diabetic and non-diabetic individuals is incompletely understood, but may involve chronic, low-grade inflammation and dysfunction of the vascular endothelium. The aim of this study was to investigate the association of inflammation and vascular endothelial dysfunction with prevalent retinopathy in both diabetic and non-diabetic individuals.
Methods A subsample of 625 individuals, stratified for age, sex and glucose tolerance status, of a population-based cohort study of 50-74 year-old-people, underwent an extensive physical and ophthalmologic examination, including funduscopy and two-field 45-degree fundus photography in mydriasis in both eyes. Plasma levels of CRP, sICAM-1, vWf, sVCAM-1, and urinary albumin-to-creatinine ratio were assessed. Summarizing Z-scores for inflammation (Z-scores of CRP and sICAM-1) and endothelial dysfunction (Z-scores of vWf, sVCAM-1 and ACR) were constructed.
Results The prevalence of retinopathy was positively associated with tertiles of CRP and sICAM-1. After adjustment for age, sex and glucose tolerance status, the highest tertile of the inflammatory Z- score was associated with retinopathy in all subjects (Odds ratio: 2.22 (1.19-4.12)). After adjustment for age and sex, the highest tertile of the endothelial dysfunction Z-score was associated with retinopathy among diabetic individuals only (Odds ratio: 4.42 (1.23-15.90). Additional adjustment for other risk factors did not materially change these associations. Mutual adjustment of the inflammatory and endothelial dysfunction Z-scores also did not change the results.
Conclusion Inflammatory activity was associated with retinopathy in subjects with and without diabetes and vascular endothelial dysfunction was associated with retinopathy in diabetic subjects only. This suggests involvement of inflammatory processes and endothelial dysfunction in the pathogenesis of retinopathy.
THE RELATIONSHIP BETWEEN INCIDENCE OF BLINDNESS DUE TO DIABETES MELLITUS AND THE QUALITY OF DIABETIC AND OPHTHALMOLOGIC CARE IN VARMIA AND MAZURY REGION , POLAND , IN YEARS 1989-2003.
El_bieta Bandurska -Stankiewicz , Dorota Wiatr, Janusz Pieczy_ski
Purpose: The aim of this study was to estimate the relationship between incidence rate (IR) of blindness due to diabetes mellitus (DM) and quality of diabetic and ophthalmologic care in Varmia and Mazury Region, Poland, in years 1989-2003.
Methods: A register of blindness in diabetic patients was established in year 1989 in Varmia and Mazury region (n=321) with the use of three independent sets of data sources. IR (new cases per 100 000 persons/year) of blindness in diabetic patients, regarding the quality of diabetic and ophthalmologic care, was estimated for 3 separate periods: 1989-1993 (no standards of diabetic care), 1994-1998 (local programs of diabetic and ophthalmologic care established and introduced, with an early administration of laser therapy), 1999-2003 (introduction of standards of combined diabetic and ophthalmologic care). The statistical analysis of the collected data was performed, with the use of Excel'97 for Windows and Statistica 5.1 programs. Results: No significant increasing and decreasing trends of IR of blindness caused by DM were found in the first two periods of observation, from 2.52 (2.40-2.64, 95%CI) in year 1989 to 5.7 (5.57-5.82) in 1997 (peak value) and to 3.08 (2.96-3.20, 95%CI) in 1998 year. From 1999 to 2003, however, significantly decreasing trends were shown in IR of blindness caused by DM: from 3.08 (2.95-3.19, 95%CI) to 0.77 (0.52-0.98, 95%CI), particularly in type 2 DM patients: from 4.66 (4.52-4.80, 95%CI) to 0.5 (0.27-0.71, 95%CI). For type 1 DM patients, decreasing tendency of blindness started earlier: from 1.42 (1.17- 1.67, 95%CI) in 1996 to 0.25 (0.04-0.47, 95%CI) in 2003; differences are statistically significant.
Conclusions: These results show decreasing IR of DM blindness in Varmia and Mazury region. Our observations confirm the significant influence of improved diabetic and ophthalmologic care with an early laser therapy program on IR of blindness in diabetic patients.
CONTACT EASDEC »