- Email: [email protected]
Identification of diabetic retinopathy by stereoscopic digital imaging via teleophthalmology: a comparison to slide film
CLINICAL STUDIES
Identification of diabetic retinopathy by stereoscopic digital imaging via teleophthalmology: a comparison to slide film Matthew T.S. Tennant, MD; Mark D.J. Greve, MD; Christopher J. Rudnisky, MD; Tim R. Hillson, MD; Brad J. Hinz, MD ABSTRACT • RESUME Background: Diabetic retinopathy is a leading cause of vision loss in North America.
We compared mydriatic seven-field stereoscopic digital imaging to I00 ASA slide film photography for the identification of diabetic retinopathy via teleophthalmology. Methods: Patients from a northern Alberta community with diabetes mellitus diagnosed by a physician were asked to participate in a teleophthalmology pilot project. Patients were enrolled at four different times between October 1999 and June 2000. Seven 30° fields of the retina were photographed with both slide film and digital imaging (resolution 2008 X 3040 pixels) through a dilated pupil. Slide film was developed and reviewed in a masked fashion by a retinal specialist. Digital images for each patient were transmitted by satellite to Edmonton and analysed a minimum of 2 months after the original slide film. Retinal abnormalities were graded with the use of the Early Treatment Diabetic Retinopathy Study extension of the modified Airlie House classification. We calculated the sensitivity and specificity of digital imaging for the identification of features of diabetic retinopathy as seen on slide film. Pearson's correlation coefficient was also calculated. Results: A total of 121 patients (241 eyes), of whom 114 (94.2%) had non-insulindependent diabetes, participated in the study. The average duration of diabetes was 8.5 years. Of the 121 patients 57 (47.1 %) had diabetic retinopathy, 12 (9.9%) had clinically significant macular edema and 2 ( 1.6%) had neovascularization. Pearson's correlation coefficient for the presence of retinopathy between slide film and stereoscopic digital imaging was 0.92 for microaneurys.ms, 0.80 for hemorrhages, 0.45 for intraretinal microvascular abnormalities, 0.32 for venous beading, 1.00 for neovascularization of the disc, 1.00 for neovascularization elsewhere in the retina and 0.97 for clinically significant macular edema (p < 0.00 I). The correlation between the two techniques for severe nonproliferative diabetic retinopathy (NPDR) was 0.86 and for high-risk proliferative diabetic retinopathy 1.00 (p < 0.00 I).
From the Department of Ophthalmology, University of Alberta, Edmonton, Alta.
Accepted for publication Apr. 10, 2001
Preliminary results of this study were presented as a poster paper at the Canadian Ophthalmological Society meeting held in Whistler, BC, June 17-20, 2000.
Reprint requests to: Dr. Mark D.J. Greve, Department of Ophthalmology, University of Alberta, Royal Alexandra Hospital, 10240 Kingsway Ave., Edmonton AB TSH 3V9; fax (780) 477-4969; mgreve@ alberta-retina.com
Originally received Sept. 18, 2000
Can j Ophthalmol 200 I;36: 187-96
Digital imaging vs. slide film-Tennant et al
187
Digital imaging vs. slide film- Tennant et al
Interpretation: Stereoscopic digital imaging has a high level of correlation with slide film for the identification of most features of diabetic retinopathy, including microaneurysms, hemorrhage, severe NPDR, highrisk proliferative diabetic retinopathy and clinically significant macular edema.
Contexte : La retinopathie diabetique est une des principales causes de Ia perte de Ia vue en Amerique du Nord. Nous avons compare l'imagerie numerique stereoscopique mydriatique a sept champs avec l'image d'une diapositive I00 ASA pour !'identification de Ia retinopathie diabetique par le biais de Ia teleophtalmologie. Methodes: Un groupe de personnes du nord de I'Alberta chez qui les medecins avaient diagnostique le diabete sucre ont ete invitees a participer a un projet pilote de teleophtalmologie. Ces patients ont ete inscrits a quatre dates entre octobre 1999 et juin 2000. On a recueilli des photographies a sept champs de 30° de Ia retine a travers Ia pupille dilatee pour obtenir des diapositives d'une part et, d'autre part, des images numeriques (resolution 2008 X 3040 pixels). Les diapositives furent developpees et examinees a l'insu par un specialiste de Ia retine. Les images numeriques de chaque patient furent transmises par satellite a Edmonton et analysees au moins deux mois apres les diapositives originales. Les anomalies de Ia retine furent classees en utilisant !'extension de Ia classification Airlie House utilisee dans Ia Early Treatment Diabetic Retinopathy Study. Nous avons calcule Ia sensibilite et Ia specificite de l'imagerie numerique pour identifier les particularites de Ia retinopathie diabetique apparaissant sur les diapositives. On a aussi calcule le coefficient de correlation de Pearson. Resultats : En tout, 121 patients (241 yeux), parmi lesquels I 14 (94,2 %) avaient un diabete non insulinodependant, ont participe a l'etude. lis avaient le diabete depuis 8,5 ans en moyenne. Parmi les 121 patients, 57 (47, I %) avaient une retinopathie diabetique, 12 (9,9 %) avaient un cedeme maculaire cliniquement important et 2 (I ,6 %) avaient une neovascularisation. Le coefficient de correlation de Pearson concernant Ia presence de Ia retinopathie entre les diapositives et l'image numerique stereoscopique etait: 0,92 pour les microanevrismes, 0,80 pour les hemorragies, 0,45 pour les anomalies microvasculaires intraretiniennes, 0,32 pour les veines en collier de perles, I,00 pour Ia neovascularisation de Ia papille, I,00 pour les autres neovascularisations de Ia retine et 0,97 pour l'cedeme maculaire cliniquement important (p < 0,00 I). La correlation entre les deux techniques etait de 0,86 pour Ia retinopathie diabetique non proliferative (NPDR) grave et de I,00 pour Ia retinopathie diabetique proliferative a risque eleve (p < 0,00 I). Interpretation : L'imagerie numerique stereoscopique a un haut degre de correlation avec les diapositives quant a !'identification de Ia plupart des caracteristiques de Ia retinopathie diabetique, notamment, le microanevrisme, l'hemorragie, le NPDR grave, Ia retinopathie diabetique proliferative a risque eleve et l'redeme maculaire cliniquement important.
D
iabetic retinopathy is a leading cause of blindness in North America. 1 In Canada diabetes mellitus is more common among people of aboriginal descent,2 with prevalence rates reported to be 2.3 times 3 to 6.7 times 4 higher than those for nonnative people. Many native people live in remote locations across northern Canada, without access to specialist
188
CAN J OPHTHALMOL-VOL. 36, NO.4, 2001
eye care. Fort Vermilion, Alta., located 640 km north of Edmonton, is the site of a teleophthalmology pilot project to improve diabetic eye care for northern Canadians. The project incorporates a novel stereoscopic digital imaging system to identify and grade levels of diabetic retinopathy. There have been major advances in digital imaging
Digital imaging vs. slide film- Tennant et al
technology in the past few years, and this technique is emerging as a possible alternative to slide film when photographing retinal disease. 5- 7 Review of the recent literature shows that digital imaging is used to screen for diabetic retinopathy. 8•9 Ryder and colleagues 10 found that digital imaging was superior to Polaroid film when screening for diabetic retinopathy. The purpose of our study differs from that of other studies, in which "screening" for diabetic retinopathy is the goal. Instead, we wanted to identify whether stereoscopic digital imaging is as sensitive and specific as seven-field slide film photography in the identification of treatable diabetic retinopathy. Patients identified by digital imaging as having treatable diabetic retinopathy were referred to a retinal specialist for treatment. All other patients were followed at a distance with digital photography repeated via teleophthalmology at a time appropriate for their level of retinopathy. Mydriatic seven-field 30° stereoscopic slide film photography of the retina is an accepted method to identify diabetic retinopathy and has been used in a number of major diabetic retinopathy epidemiology and treatment trials. 1•11 - 13 It has been found to be a cost-effective alternative to clinical examination for people living in remote communities. 14 Although contact lens biomicroscopy may be the most sensitive method to identify clinically significant macular edema, slide film photography is both reliable and reproducible. 15 We compared seven-field 30° stereoscopic digital imaging to slide film photography for the identification of diabetic retinopathy in patients with diabetes living in and around Fort Vermilion. To our knowledge, this study represents the first published report regarding the identification of treatable diabetic retinopathy with high-resolution (2008 X 3040 pixels) stereoscopic digital imaging. METHODS
Patients with diabetes from Fort Vermilion and surrounding communities were asked to participate in a teleophthalmology pilot project between Fort Vermilion and Edmonton. Patients were eligible for inclusion in the study if they had diabetes diagnosed by a physician. Patients were enrolled without regard for age, sex, duration of diabetes, previous ophthalmic surgery, laser procedures, clarity of ocular media, level of diabetic retinopathy or presence of other eye disease. Local physicians and community health representatives organized all patient recruitment and clinic appointment bookings. The study was approved by
the Ethics Committee of the Faculty of Medicine, University of Alberta, Edmonton. Patients were enrolled at four different times between October 1999 and June 2000. After informed consent was obtained, the visual acuity and intraocular pressure were recorded by a trained examiner. Eyes were dilated with a combination of 2.5% phenylephrine hydrochloride and 1.0% tropicarnide (Diophenyl-T, Dioptic Laboratories, Markham, Ont.) 30 minutes before retinal photography by a certified ophthalmic photographer. Seven 30° fields of the retina were photographed with a fundus camera (Zeiss FF450, Carl Zeiss Corporation, Jena, Germany). 16 Two nonsimultaneous sets of images were captured for each eye, one with a slide film camera (F3, Nikon Corporation, Tokyo) onto 100 ASA film (Elitechrome, Eastman Kodak, Rochester, NY) and one with a digital camera (DCS 560, Eastman Kodak [resolution 2008 X 3040 pixels]). Digital and slide film nonsimultaneous stereoscopic pair photographs were captured of the disc and macula (field 1 and 2) through a corneal-induced parallaxY Nonstereoscopic images were captured of fields 3 to 7. The digital images were transferred to a laptop computer for review, compression and transfer by satellite to Edmonton. The slide film photographs were shipped to Edmonton for development. A retinal specialist reviewed all slide film photographs within 1 week of capture. Stereoscopic film was analysed with a +5-dioptre stereo viewer on a light table in a masked fashion. Nonstereoscopic fields 3 to 7 were viewed with a + 10-dioptre viewing lens. Left and right eyes were analysed on separate days to minimize intereye bias. All images (digital and 35-mm) were reviewed for the presence or absence of microaneurysms, intraretinal hemorrhage, cotton-wool spots, hard exudate, venous beading, intraretinal microvascular abnormalities (IRMAs), macular edema, clinically significant macular edema, neovascularization of the disc, neovascularization elsewhere in the retina, fibrous proliferation, focal laser scars and panretinal photocoagulation scars. Retinal abnormalities were graded with the use of the Early Treatment Diabetic Retinopathy Study (ETDRS) extension of the modified Airlie House classification (Table 1). 16 Standard diabetic retinal photographs (SPs) were supplied by the ETDRS reading centre (Table 1). Digital images for each patient were analysed in a masked fashion a minimum of 2 months after the slide film was reviewed. Again, left and right eyes were analysed on separate days. Images were displayed on a computer screen with
CAN
J OPHTHALMOL-VOL.
36, NO.4, 2001
189
Digital imaging vs. slide film- Tennant et al
Table 1-Modified Early Treatment Diabetic Retinopathy Study grading 11 of retinal abnormalities 16 with associated standard photographs (SPs) Abnormality
Field assessed
Microaneurysms Hemorrhage
1-7 2-7
Nerve fibre layer infarcts
1-7
Hard exudate
2-7
Intrareti nal microvascular abnormalities
2-7
Venous beading
3-7
Venous loops
3-7
Neovascularization of disc
Neovascu larization elsewhere
2-7
16-bit colour and resolution capability of 1024 X 768 pixels. Images were scaled automatically so the entire photograph could be viewed on the screen at once. The ability to magnify the image was at the discretion of the reader. Stereoscopic digital images were viewed with liquid crystal display shutter goggles (Stereographics Corp., San Rafael, Calif.) and a highdefinition video monitor. The digital images were not enhanced or altered in any way before or during evaluation. We calculated the sensitivity and specificity of stereoscopic digital imaging for identifying all features associated with both nonproliferative diabetic retinopathy (NPDR) and proliferative diabetic reti-
190
CAN J OPHTHALMOL-VOL. 36, NO. 4, 200 I
Grading None, I MA, :2:2MA None < SPI Definite :2: SPI, < SP2A Moderate :2: SP2A, < SP2B Severe :2: SP2B Very severe None < SPSA Definite :2: SPSA, < SPS Moderate :2: SPS Severe None < SP3 Definite :2: SP3, < SPS Moderate :2: SP 5, < SP4 Severe :2: SPS severe Very None < SPSA Definite SPSA, < SPSB :2: Moderate SPSB :2: Severe None < SP6A Definite :2: SP6A, < SP6B Moderate :2: SP6B Severe None Present None < SPIOA Definite :2: SPIOA, < SPIOC Moderate :2:SPIOC Severe None > 1/ 2 disc area Definite
nopathy as seen on slide film. Pearson's correlation coefficient, a statistic that describes the degree to which two variables correspond to one another, 18 was also calculated. Analysis was first performed for the presence or absence of features associated with diabetic retinopathy. Once complete, the data were then analysed for correlation between digital and film imaging when identifying treatable diabetic retinopathy. Treatable disease was defined as one or more of the following: severe NPDR, high-risk proliferative diabetic retinopathy and clinically significant macular edema. Severe NPDR was included as a treatable disease, as panretinal photocoagulation may be of benefit for older patients with non-
Digital imaging vs. slide film- Tennant et al
insulin-dependent diabetes with this level of retinopathy.19 Severe NPDR was defined as at least one of the following: four fields of severe hemorrhage (SP2A or greater), two fields of definite venous beading (SP6A or greater) or one field of moderate IRMAs (SP8A or greater). 20 High-risk proliferative diabetic retinopathy was defined as any three of the following features: neovascularization; neovascularization of the disc; neovascularization of the disc greater than SPlOA or neovascularization elsewhere greater than 1/2 disc area; and vitreous or preretinal hemorrhage. 21 Clinically significant macular edema was defined as thickening of the retina at or within 500 J.llll of the centre of the macula; hard exudates at or within 500 J.llll of the centre of the macula, if associated with thickening of adjacent retina; and a zone or zones of retinal thickening I disc area or larger, any part of which is within 1 disc diameter of the centre of the macula. 22 RESULTS
A total of 121 patients (241 eyes and I prosthesis) participated in the study. There were 69 female subjects (57.0%) and 52 male subjects (43.0%). The average age was 57.0 (range 9.0 to 83.2) years. The average duration of diabetes since diagnosis was 8.5 years (range 2 weeks to 40.0 years). Of the 121 patients 114 (94.2%) had non-insulin-dependent diabetes and 7 (5.8%) had insulin-
dependent diabetes. All patients enrolled at the beginning of the study were included in the final analysis. Of the 12I patients 57 (47.1 %) had diabetic retinopathy when evaluated with slide film. Twelve patients (9.9%) were identified with treatable diabetic retinopathy. Of the 12, 11 (9 .I%) were identified with clinically significant macular edema in at least one eye, and 1 (0.8%) was identified with clinically significant macular edema, four fields of severe hemorrhage and neovascularization of the disc. Correlation by eye between stereoscopic digital imaging and slide film photography varied for features associated with NPDR (Table 2). Pearson's correlation coefficient ranged from 0.92 (p < 0.001) when identifying microaneurysms to 0.32 (p < 0.001) when identifying venous beading. When identifying proliferative diabetic retinopathy, Pearson's correlation coefficient ranged from 1.00 (p < 0.001) for neovascularization of the disc and neovascularization elsewhere to 0.58 (p < 0.001) for fibrous proliferation. No vitreous hemorrhage or preretinal hemorrhage was identified. Clinically significant macular edema was identified in 16 eyes by digital imaging and 17 eyes by slide film (Pearson's correlation coefficient 0.97 [p
Table 2-Correlation between digital imaging and slide film in the identification of any level of diabetic retinopathy
Diabetic retinopathy feature Venous beading lntraretinal microvascular abnormalities Fibrous proliferation Focal laser scar Nerve fibre layer infarcts Hard exudate Hemorrhage Macular edema Microaneurysm Clinically significant macular edema Panretinal photocoagulation scars Neovascularization of disc Neovascularization elsewhere *Two-tailed, n
No. of eyes as identified with slide film
Pearson's correlation coefficient*
5
0.32 0.45 0.58 0.62 0.68 0.72 0.80 0.84 0.92 0.97 1.00 1.00 1.00
II 3 7 27 37 58 29 86 17 7 2 2
= 241, p < 0.00 I.
CAN J OPHTHALMOL-VOL. 36, NO. 4, 200 I
I 91
Digital imaging vs. slide film- Tennant et al
Table 3-Sensitivity and specificity of stereoscopic digital imaging in the identification of diabetic retinopathy as seen with slide film Diabetic retinopathy feature Venous beading lntraretinal microvascular abnormalities Fibrous proliferation Focal laser scar Nerve fibre layer infarcts Hard exudate Hemorrhage Macular edema Microaneurysm Clinically significant macular edema Panretinal photocoagulation scars Neovascularization of disc Neovascularization elsewhere
Sensitivity, %
Specificity, %
40.0
28.6
45.5 33.3 71.4 70.4 83.8 86.2 82.9 95.3
50.0 100.0 55.6 73.1 70.5 83.3 88.9 94.3
94.1
100.0
100.0 100.0 100.0
100.0 100.0 100.0
relation coefficient 0.86 [p < 0.001]) (Table 4). Severe hemorrhage (SP2A or greater) was identified in two eyes with both techniques. Venous beading greater than SP6A was found in one eye with digital imaging. No eyes were identified with venous beading greater than SP6A with slide film. IRMAs greater than SP8A were identified in one eye with digital imaging and one
neovascularization of the disc (Table 3). The specificity of digital imaging ranged from 28.6% for venous beading to 100.0% for neovascularization of the disc. No venous loops were identified with either slide film or digital imaging. Severe NPDR was identified in three eyes with slide film and four eyes with digital imaging (Pearson's cor-
Table 4-Correlation between stereoscopic digital imaging and slide film in the identification of severe nonproliferative diabetic retinopathy No. of eyes Diabetic retinopathy feature Four fields of severe hemorrhage Two fields of venous beading One field of intraretinal microvascular abnormalities Severe nonproliferative diabetic retinopathy *Two-tailed, n
192
= 241, p < 0.00 I.
CAN J OPHTHALMOL-VOL. 36, NO.4, 2001
Slide film
Digital imaging
Pearson's correlation coefficient*
2
2
1.00 0.00
0
0.00
3
4
0.86
Digital imaging vs. slide film- Tennant et al
different eye with slide film. In both instances the eye identified with IRMAs greater than SP8A with one imaging technique was identified as having IRMAs less than SP8A with the other technique. High-risk proliferative diabetic retinopathy was identified in four eyes with both digital imaging and slide film. Neovascularization of the disc was found in two eyes (less than SPlOA in one eye and greater than SPlOA in one eye). Neovascularization elsewhere was identified in two eyes with both techniques (less than 1/ disc area in one eye and greater than '1 disc area in 2 2 one eye). No eyes were identified with preretinal hemorrhage or vitreous hemorrhage. One eye was identified with three of four high-risk characteristics with both imaging techniques (Pearson's correlation coefficient 1.00 [p < 0.001]). INTERPRETATION
Diabetic retinopathy was present in about half (47.1%) of all patients assessed as part of this pilot study. Although the sample population was not selected randomly, and enrolment bias may have been present, our study does provide an estimate of the current level of diabetic retinopathy in a largely aboriginal population in northern Alberta. In addition, the proportion of patients with treatable disease, 9.9%, is consistent with rates reported in other epidemiologic studies. 13 ·23 Identification of microaneurysms, the cornerstone of early diabetic retinopathy, was similar with slide film and digital imaging, with a correlation coefficient of 0.92. The identification of nonproliferative disorders, including hemorrhage, nerve fibre layer infarcts and hard exudate, also correlated well with slide film. This demonstrates the potential of high-resolution digital imaging to identify early diabetic retinopathy (Fig. 1, top). The correlation coefficient for identifying severe NPDR was 0.86. Correlation was less when identifying IRMAs (0.45) and venous beading (0.32). Following the masked comparison of digital and slide film photographs, discrepancies were identified. Discrepancies were then correlated to individual differences between each digital and film image in an unmasked fashion. The poor correlation between slide film and digital imaging for IRMAs was attributed to two factors. The areas of IRMAs were small and subtle, which led to "missed" identification if image detail was not perfect (Fig. 1, bottom). In addition, the resolution of digital imaging (2008 X 3040 pixels)
was less than with slide film. This level of resolution led to difficulties in identifying IRMAs with certainty. The poor correlation between slide film and digital imaging in the identification of venous beading was attributed to the very low incidence and severity of venous beading in the patients included in the study. In fact, only one eye was found to have venous beading greater than SP6A with either slide film or digital imaging. Lim and associates 9 recently reported the results of a comparison between a nonmydriatic digital camera (resolution 640 X 480 pixels) and slide film. They found low sensitivity for almost all features of diabetic retinopathy, including microaneurysms (46%), dot or blot hemorrhages (71%), flame (50%), hard exudates (64% ), nerve fibre layer infarcts (50%), clinically significant macular edema (30% ), venous beading (50%), neovascularization of the disc (25%) and neovascularization elsewhere in the retina (29% ). The sensitivity of digital imaging in identifying IRMAs (found in one eye) was 100%. Specificity values were much higher than sensitivity values, with many being greater than 90%. We used a digital camera with much greater resolution and therefore had higher sensitivity values for all features except IRMAs and venous beading. In our study identification of neovascularization of the disc (two eyes) and neovascularization elsewhere (two eyes) (Fig. 2, top) was identical with slide film and digital imaging. One patient had three of four high-risk characteristics and was referred to a retinal specialist for treatment. Although the correlation between slide film and digital imaging for severe NPDR and proliferative disease was excellent, only two patients had retinopathy this severe. Interpretation of these findings is therefore limited. Stereoscopic digital imaging had 94.1% sensitivity and 100% specificity in identifying clinically significant macular edema (Fig. 2, bottom). All patients identified with clinically significant macular edema with slide film were referred to a tertiary centre for focal laser treatment. Contact lens biomicroscopy was performed on each eye, confirming clinically significant macular edema in each case. Stereoscopic digital imaging did not identify one eye determined to have clinically significant macular edema on slide film. Following completion of the masked study this image pair was unmasked and reanalysed. It was discovered that although both imaging techniques provided relatively poor detail of the
CAN J OPHTHALMOL-VOL. 36, NO.4, 2001
193
Digital imaging vs. slide film- Tennant et al
Fig. 1-Top left: Fundus photograph obtained with digital camera from patient with diabetic retinopathy, showing microaneurysms. Top right: Magnified view of area of interest. Bottom left: Digital image, showing intraretinal microvascular abnormalities. Bottom right: Magnified view of area of interest.
retina owing to early cataract and poor image focus, the digital image pair was of inferior quality. Even with repeat image analysis clinically significant macular edema could not be identified with any degree of certainty when viewing the digital images. Fortunately, this patient had been identified by digital imaging as having clinically significant macular edema in the fellow eye and would therefore have been referred for treatment by a retinal specialist.
194
CAN
J OPHTHALMOL-VOL. 36, NO. 4, 200 I
CONCLUSION
Stereoscopic digital imaging is both sensitive and specific in identifying diabetic retinopathy as seen on slide film. There is a high degree of correlation between slide film and digital imaging when identifying the presence of rnicroaneurysms, hemorrhage and clinically significant macular edema. Despite the very low incidence of neovascularization of the disc,
Digital imaging vs. slide film- Tennant et al
1-ig. 2-1 op lett: LJigital image, showing neovascularization elsewhere in retina (other than ot disc). I op right: Magnitied view ot area of interest. Bottom left: Digital image, showing clinically significant macular edema. Bottom right: Magnified view of area of interest.
neovascularization elsewhere in the retina and severe NPDR, the correlation between slide film and digital imaging for identifying these features was high. This novel technology may play a role in the screening and diagnosis of treatable diabetic retinopathy via teleophthalmology. This study was supported by the University of Alberta, Edmonton, MacDonald Dettwiler and Associates, Vancouver, Precam and Associates, Ottawa, and the Canadian Space Agency, Ottawa. None of the authors hold any proprietary interests in any of the products mentioned in this article.
REFERENCES
1. Diabetic Retinopathy Study Research Group. Preliminary
report on effects of photocoagulation therapy. Am J Ophthalmo/1976;94:505-37. 2. MacMillan HL, MacMillan AB, Offord DR, Dingle JL. Aboriginal health. Can Med Assoc J 1996;155: 1569-77. 3. Young TK, Shah C. Extent and magnitude of the problem. In: Young TK, editor. Diabetes in the Canadian Native population: bicultural perspectives. Toronto: Canadian Diabetes Association; 1987. p. 11-25. 4. Evers S, McCracken E, Antone T, Deagle G. Prevalence of diabetes in Indians and Caucasians living in southwestern Ontario. Can J Public Health 1987;78:240-3.
CAN J OPHTHALMOL-VOL. 36, NO.4, 200 I
195
Digital imaging vs. slide film- Tennant et al
5. McGregor G. Transmitting digital images of the fundus for the West Virginia Diabetic Eye Care Project. J Ophthalmic Photogr 1994;16:87-98. 6. YoungS, George LD, Lusty J, Owens DR. A new screening tool for diabetic retinopathy: the Canon CR5 45 NM retinal camera with Frost Medical Software RIS-Lite digital imaging system. J Audiov Media Med 1997;20:11-4. 7. Li HK. Telemedicine and ophthalmology. Surv Ophthalmol 1999;44:61-72. 8. LinDY, Blumenkranz MS, Brothers R. The role of digital fundus photography in diabetic retinopathy screening. Diabetes Techno[ Ther 1999;1:477-87. 9. Lim JI, LaBree L, Nichols T, Cardenas I. A comparison of digital nonmydriatic fundus imaging with standard 35millimeter slides for diabetic retinopathy. Ophthalmology 2000;107:866-76. 10. Ryder REJ, Kong N, Bates AS, Sim J, Welch J, Kritzinger EE. Instant electronic imaging systems are superior to Polaroid at detecting sight-threatening diabetic retinopathy. Diabetic Med 1998;15:254--8. 11. Diabetic Retinopathy Study Group. Report 7: a modification of the Airlie House classification of diabetic retinopathy./nvest Ophthalmol Vis Sci 1981;21:210-26. 12. Klein R, Klein BEK, Moss SE, Davis MD, DeMets DL. The Wisconsin Epidemiologic Study of Diabetic Retinopathy. IX. Four-year incidence and progression of diabetic retinopathy when age at diagnosis is less than 30 years. Arch Ophthalmoll989;l07:237-43. 13. Mitchell P, Smith W, Wang JJ, Attebo K. Prevalence of diabetic retinopathy in an older community. The Blue Mountains Eye Study. Ophthalmology 1998;105:406-11. 14. Martin JD, Yidegiligne HM. The cost-effectiveness of a retinal photography screening program for preventing diabetic retinopathy in the First Nations diabetic population in British Columbia, Canada./nt J Circumpolar Health 1998; 57(suppl1):379-82.
196
CAN J OPHTHALMOL-VOL. 36, NO. 4, 200 I
15. Early Treatment Diabetic Retinopathy Study Research Group. Detection of diabetic macular edema: ophthalmoscopy versus photography -Early Treatment Diabetic Retinopathy Study report #5. Ophthalmology 1989;96: 746-51. 16. Early Treatment Diabetic Retinopathy Study Research Group. Grading diabetic retinopathy from stereoscopic color fundus photographs - an extension of the modified Airlie House classification. ETDRS report number 10. Ophthalmology 1991;98(5 suppl):786-806. 17. Wong D. Textbook of ophthalmic photography. New York: Inter-Optics Publications; 1982. p. 76. 18. Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical epidemiology: a basic science for clinical medicine. Toronto: Little, Brown and Company; 1991. p. 69-152. 19. Ferris F. Early photocoagulation in patients with either type I or type II diabetes. Trans Am Ophthalmol Soc 1996;94:505-37. 20. Early Treatment Diabetic Retinopathy Study Research Group. Fundus photographic risk factors for progression of diabetic retinopathy. ETDRS report number 12. Ophthalmology 1991;98:823-33. 21. Diabetic Retinopathy Study Research Group. Four risk factors for severe visual loss in diabetic retinopathy. Arch Ophthalmoll979;97:654-5. 22. Early Treatment Diabetic Retinopathy Study Research Group. Photocoagulation of macular edema. Early Treatment Diabetic Study report #1. Arch Ophthalmol 1985; 103:1796-806. 23. Leske MC, Wu SY, Hyman L, Li X, Hennis A, Connell AM, et al. Diabetic retinopathy in a black population: the Barbados Eye Study. Ophthalmology 1999;106:1893-9.
Key words: diabetic retinopathy, screening, teleophthalmology, imaging techniques, digital stereoscopic photography
Identification of diabetic retinopathy by stereoscopic digital imaging via teleophthalmology: a comparison to slide film Matthew T.S. Tennant, MD; Mark D.J. Greve, MD; Christopher J. Rudnisky, MD; Tim R. Hillson, MD; Brad J. Hinz, MD ABSTRACT • RESUME Background: Diabetic retinopathy is a leading cause of vision loss in North America.
We compared mydriatic seven-field stereoscopic digital imaging to I00 ASA slide film photography for the identification of diabetic retinopathy via teleophthalmology. Methods: Patients from a northern Alberta community with diabetes mellitus diagnosed by a physician were asked to participate in a teleophthalmology pilot project. Patients were enrolled at four different times between October 1999 and June 2000. Seven 30° fields of the retina were photographed with both slide film and digital imaging (resolution 2008 X 3040 pixels) through a dilated pupil. Slide film was developed and reviewed in a masked fashion by a retinal specialist. Digital images for each patient were transmitted by satellite to Edmonton and analysed a minimum of 2 months after the original slide film. Retinal abnormalities were graded with the use of the Early Treatment Diabetic Retinopathy Study extension of the modified Airlie House classification. We calculated the sensitivity and specificity of digital imaging for the identification of features of diabetic retinopathy as seen on slide film. Pearson's correlation coefficient was also calculated. Results: A total of 121 patients (241 eyes), of whom 114 (94.2%) had non-insulindependent diabetes, participated in the study. The average duration of diabetes was 8.5 years. Of the 121 patients 57 (47.1 %) had diabetic retinopathy, 12 (9.9%) had clinically significant macular edema and 2 ( 1.6%) had neovascularization. Pearson's correlation coefficient for the presence of retinopathy between slide film and stereoscopic digital imaging was 0.92 for microaneurys.ms, 0.80 for hemorrhages, 0.45 for intraretinal microvascular abnormalities, 0.32 for venous beading, 1.00 for neovascularization of the disc, 1.00 for neovascularization elsewhere in the retina and 0.97 for clinically significant macular edema (p < 0.00 I). The correlation between the two techniques for severe nonproliferative diabetic retinopathy (NPDR) was 0.86 and for high-risk proliferative diabetic retinopathy 1.00 (p < 0.00 I).
From the Department of Ophthalmology, University of Alberta, Edmonton, Alta.
Accepted for publication Apr. 10, 2001
Preliminary results of this study were presented as a poster paper at the Canadian Ophthalmological Society meeting held in Whistler, BC, June 17-20, 2000.
Reprint requests to: Dr. Mark D.J. Greve, Department of Ophthalmology, University of Alberta, Royal Alexandra Hospital, 10240 Kingsway Ave., Edmonton AB TSH 3V9; fax (780) 477-4969; mgreve@ alberta-retina.com
Originally received Sept. 18, 2000
Can j Ophthalmol 200 I;36: 187-96
Digital imaging vs. slide film-Tennant et al
187
Digital imaging vs. slide film- Tennant et al
Interpretation: Stereoscopic digital imaging has a high level of correlation with slide film for the identification of most features of diabetic retinopathy, including microaneurysms, hemorrhage, severe NPDR, highrisk proliferative diabetic retinopathy and clinically significant macular edema.
Contexte : La retinopathie diabetique est une des principales causes de Ia perte de Ia vue en Amerique du Nord. Nous avons compare l'imagerie numerique stereoscopique mydriatique a sept champs avec l'image d'une diapositive I00 ASA pour !'identification de Ia retinopathie diabetique par le biais de Ia teleophtalmologie. Methodes: Un groupe de personnes du nord de I'Alberta chez qui les medecins avaient diagnostique le diabete sucre ont ete invitees a participer a un projet pilote de teleophtalmologie. Ces patients ont ete inscrits a quatre dates entre octobre 1999 et juin 2000. On a recueilli des photographies a sept champs de 30° de Ia retine a travers Ia pupille dilatee pour obtenir des diapositives d'une part et, d'autre part, des images numeriques (resolution 2008 X 3040 pixels). Les diapositives furent developpees et examinees a l'insu par un specialiste de Ia retine. Les images numeriques de chaque patient furent transmises par satellite a Edmonton et analysees au moins deux mois apres les diapositives originales. Les anomalies de Ia retine furent classees en utilisant !'extension de Ia classification Airlie House utilisee dans Ia Early Treatment Diabetic Retinopathy Study. Nous avons calcule Ia sensibilite et Ia specificite de l'imagerie numerique pour identifier les particularites de Ia retinopathie diabetique apparaissant sur les diapositives. On a aussi calcule le coefficient de correlation de Pearson. Resultats : En tout, 121 patients (241 yeux), parmi lesquels I 14 (94,2 %) avaient un diabete non insulinodependant, ont participe a l'etude. lis avaient le diabete depuis 8,5 ans en moyenne. Parmi les 121 patients, 57 (47, I %) avaient une retinopathie diabetique, 12 (9,9 %) avaient un cedeme maculaire cliniquement important et 2 (I ,6 %) avaient une neovascularisation. Le coefficient de correlation de Pearson concernant Ia presence de Ia retinopathie entre les diapositives et l'image numerique stereoscopique etait: 0,92 pour les microanevrismes, 0,80 pour les hemorragies, 0,45 pour les anomalies microvasculaires intraretiniennes, 0,32 pour les veines en collier de perles, I,00 pour Ia neovascularisation de Ia papille, I,00 pour les autres neovascularisations de Ia retine et 0,97 pour l'cedeme maculaire cliniquement important (p < 0,00 I). La correlation entre les deux techniques etait de 0,86 pour Ia retinopathie diabetique non proliferative (NPDR) grave et de I,00 pour Ia retinopathie diabetique proliferative a risque eleve (p < 0,00 I). Interpretation : L'imagerie numerique stereoscopique a un haut degre de correlation avec les diapositives quant a !'identification de Ia plupart des caracteristiques de Ia retinopathie diabetique, notamment, le microanevrisme, l'hemorragie, le NPDR grave, Ia retinopathie diabetique proliferative a risque eleve et l'redeme maculaire cliniquement important.
D
iabetic retinopathy is a leading cause of blindness in North America. 1 In Canada diabetes mellitus is more common among people of aboriginal descent,2 with prevalence rates reported to be 2.3 times 3 to 6.7 times 4 higher than those for nonnative people. Many native people live in remote locations across northern Canada, without access to specialist
188
CAN J OPHTHALMOL-VOL. 36, NO.4, 2001
eye care. Fort Vermilion, Alta., located 640 km north of Edmonton, is the site of a teleophthalmology pilot project to improve diabetic eye care for northern Canadians. The project incorporates a novel stereoscopic digital imaging system to identify and grade levels of diabetic retinopathy. There have been major advances in digital imaging
Digital imaging vs. slide film- Tennant et al
technology in the past few years, and this technique is emerging as a possible alternative to slide film when photographing retinal disease. 5- 7 Review of the recent literature shows that digital imaging is used to screen for diabetic retinopathy. 8•9 Ryder and colleagues 10 found that digital imaging was superior to Polaroid film when screening for diabetic retinopathy. The purpose of our study differs from that of other studies, in which "screening" for diabetic retinopathy is the goal. Instead, we wanted to identify whether stereoscopic digital imaging is as sensitive and specific as seven-field slide film photography in the identification of treatable diabetic retinopathy. Patients identified by digital imaging as having treatable diabetic retinopathy were referred to a retinal specialist for treatment. All other patients were followed at a distance with digital photography repeated via teleophthalmology at a time appropriate for their level of retinopathy. Mydriatic seven-field 30° stereoscopic slide film photography of the retina is an accepted method to identify diabetic retinopathy and has been used in a number of major diabetic retinopathy epidemiology and treatment trials. 1•11 - 13 It has been found to be a cost-effective alternative to clinical examination for people living in remote communities. 14 Although contact lens biomicroscopy may be the most sensitive method to identify clinically significant macular edema, slide film photography is both reliable and reproducible. 15 We compared seven-field 30° stereoscopic digital imaging to slide film photography for the identification of diabetic retinopathy in patients with diabetes living in and around Fort Vermilion. To our knowledge, this study represents the first published report regarding the identification of treatable diabetic retinopathy with high-resolution (2008 X 3040 pixels) stereoscopic digital imaging. METHODS
Patients with diabetes from Fort Vermilion and surrounding communities were asked to participate in a teleophthalmology pilot project between Fort Vermilion and Edmonton. Patients were eligible for inclusion in the study if they had diabetes diagnosed by a physician. Patients were enrolled without regard for age, sex, duration of diabetes, previous ophthalmic surgery, laser procedures, clarity of ocular media, level of diabetic retinopathy or presence of other eye disease. Local physicians and community health representatives organized all patient recruitment and clinic appointment bookings. The study was approved by
the Ethics Committee of the Faculty of Medicine, University of Alberta, Edmonton. Patients were enrolled at four different times between October 1999 and June 2000. After informed consent was obtained, the visual acuity and intraocular pressure were recorded by a trained examiner. Eyes were dilated with a combination of 2.5% phenylephrine hydrochloride and 1.0% tropicarnide (Diophenyl-T, Dioptic Laboratories, Markham, Ont.) 30 minutes before retinal photography by a certified ophthalmic photographer. Seven 30° fields of the retina were photographed with a fundus camera (Zeiss FF450, Carl Zeiss Corporation, Jena, Germany). 16 Two nonsimultaneous sets of images were captured for each eye, one with a slide film camera (F3, Nikon Corporation, Tokyo) onto 100 ASA film (Elitechrome, Eastman Kodak, Rochester, NY) and one with a digital camera (DCS 560, Eastman Kodak [resolution 2008 X 3040 pixels]). Digital and slide film nonsimultaneous stereoscopic pair photographs were captured of the disc and macula (field 1 and 2) through a corneal-induced parallaxY Nonstereoscopic images were captured of fields 3 to 7. The digital images were transferred to a laptop computer for review, compression and transfer by satellite to Edmonton. The slide film photographs were shipped to Edmonton for development. A retinal specialist reviewed all slide film photographs within 1 week of capture. Stereoscopic film was analysed with a +5-dioptre stereo viewer on a light table in a masked fashion. Nonstereoscopic fields 3 to 7 were viewed with a + 10-dioptre viewing lens. Left and right eyes were analysed on separate days to minimize intereye bias. All images (digital and 35-mm) were reviewed for the presence or absence of microaneurysms, intraretinal hemorrhage, cotton-wool spots, hard exudate, venous beading, intraretinal microvascular abnormalities (IRMAs), macular edema, clinically significant macular edema, neovascularization of the disc, neovascularization elsewhere in the retina, fibrous proliferation, focal laser scars and panretinal photocoagulation scars. Retinal abnormalities were graded with the use of the Early Treatment Diabetic Retinopathy Study (ETDRS) extension of the modified Airlie House classification (Table 1). 16 Standard diabetic retinal photographs (SPs) were supplied by the ETDRS reading centre (Table 1). Digital images for each patient were analysed in a masked fashion a minimum of 2 months after the slide film was reviewed. Again, left and right eyes were analysed on separate days. Images were displayed on a computer screen with
CAN
J OPHTHALMOL-VOL.
36, NO.4, 2001
189
Digital imaging vs. slide film- Tennant et al
Table 1-Modified Early Treatment Diabetic Retinopathy Study grading 11 of retinal abnormalities 16 with associated standard photographs (SPs) Abnormality
Field assessed
Microaneurysms Hemorrhage
1-7 2-7
Nerve fibre layer infarcts
1-7
Hard exudate
2-7
Intrareti nal microvascular abnormalities
2-7
Venous beading
3-7
Venous loops
3-7
Neovascularization of disc
Neovascu larization elsewhere
2-7
16-bit colour and resolution capability of 1024 X 768 pixels. Images were scaled automatically so the entire photograph could be viewed on the screen at once. The ability to magnify the image was at the discretion of the reader. Stereoscopic digital images were viewed with liquid crystal display shutter goggles (Stereographics Corp., San Rafael, Calif.) and a highdefinition video monitor. The digital images were not enhanced or altered in any way before or during evaluation. We calculated the sensitivity and specificity of stereoscopic digital imaging for identifying all features associated with both nonproliferative diabetic retinopathy (NPDR) and proliferative diabetic reti-
190
CAN J OPHTHALMOL-VOL. 36, NO. 4, 200 I
Grading None, I MA, :2:2MA None < SPI Definite :2: SPI, < SP2A Moderate :2: SP2A, < SP2B Severe :2: SP2B Very severe None < SPSA Definite :2: SPSA, < SPS Moderate :2: SPS Severe None < SP3 Definite :2: SP3, < SPS Moderate :2: SP 5, < SP4 Severe :2: SPS severe Very None < SPSA Definite SPSA, < SPSB :2: Moderate SPSB :2: Severe None < SP6A Definite :2: SP6A, < SP6B Moderate :2: SP6B Severe None Present None < SPIOA Definite :2: SPIOA, < SPIOC Moderate :2:SPIOC Severe None > 1/ 2 disc area Definite
nopathy as seen on slide film. Pearson's correlation coefficient, a statistic that describes the degree to which two variables correspond to one another, 18 was also calculated. Analysis was first performed for the presence or absence of features associated with diabetic retinopathy. Once complete, the data were then analysed for correlation between digital and film imaging when identifying treatable diabetic retinopathy. Treatable disease was defined as one or more of the following: severe NPDR, high-risk proliferative diabetic retinopathy and clinically significant macular edema. Severe NPDR was included as a treatable disease, as panretinal photocoagulation may be of benefit for older patients with non-
Digital imaging vs. slide film- Tennant et al
insulin-dependent diabetes with this level of retinopathy.19 Severe NPDR was defined as at least one of the following: four fields of severe hemorrhage (SP2A or greater), two fields of definite venous beading (SP6A or greater) or one field of moderate IRMAs (SP8A or greater). 20 High-risk proliferative diabetic retinopathy was defined as any three of the following features: neovascularization; neovascularization of the disc; neovascularization of the disc greater than SPlOA or neovascularization elsewhere greater than 1/2 disc area; and vitreous or preretinal hemorrhage. 21 Clinically significant macular edema was defined as thickening of the retina at or within 500 J.llll of the centre of the macula; hard exudates at or within 500 J.llll of the centre of the macula, if associated with thickening of adjacent retina; and a zone or zones of retinal thickening I disc area or larger, any part of which is within 1 disc diameter of the centre of the macula. 22 RESULTS
A total of 121 patients (241 eyes and I prosthesis) participated in the study. There were 69 female subjects (57.0%) and 52 male subjects (43.0%). The average age was 57.0 (range 9.0 to 83.2) years. The average duration of diabetes since diagnosis was 8.5 years (range 2 weeks to 40.0 years). Of the 121 patients 114 (94.2%) had non-insulin-dependent diabetes and 7 (5.8%) had insulin-
dependent diabetes. All patients enrolled at the beginning of the study were included in the final analysis. Of the 12I patients 57 (47.1 %) had diabetic retinopathy when evaluated with slide film. Twelve patients (9.9%) were identified with treatable diabetic retinopathy. Of the 12, 11 (9 .I%) were identified with clinically significant macular edema in at least one eye, and 1 (0.8%) was identified with clinically significant macular edema, four fields of severe hemorrhage and neovascularization of the disc. Correlation by eye between stereoscopic digital imaging and slide film photography varied for features associated with NPDR (Table 2). Pearson's correlation coefficient ranged from 0.92 (p < 0.001) when identifying microaneurysms to 0.32 (p < 0.001) when identifying venous beading. When identifying proliferative diabetic retinopathy, Pearson's correlation coefficient ranged from 1.00 (p < 0.001) for neovascularization of the disc and neovascularization elsewhere to 0.58 (p < 0.001) for fibrous proliferation. No vitreous hemorrhage or preretinal hemorrhage was identified. Clinically significant macular edema was identified in 16 eyes by digital imaging and 17 eyes by slide film (Pearson's correlation coefficient 0.97 [p
Table 2-Correlation between digital imaging and slide film in the identification of any level of diabetic retinopathy
Diabetic retinopathy feature Venous beading lntraretinal microvascular abnormalities Fibrous proliferation Focal laser scar Nerve fibre layer infarcts Hard exudate Hemorrhage Macular edema Microaneurysm Clinically significant macular edema Panretinal photocoagulation scars Neovascularization of disc Neovascularization elsewhere *Two-tailed, n
No. of eyes as identified with slide film
Pearson's correlation coefficient*
5
0.32 0.45 0.58 0.62 0.68 0.72 0.80 0.84 0.92 0.97 1.00 1.00 1.00
II 3 7 27 37 58 29 86 17 7 2 2
= 241, p < 0.00 I.
CAN J OPHTHALMOL-VOL. 36, NO. 4, 200 I
I 91
Digital imaging vs. slide film- Tennant et al
Table 3-Sensitivity and specificity of stereoscopic digital imaging in the identification of diabetic retinopathy as seen with slide film Diabetic retinopathy feature Venous beading lntraretinal microvascular abnormalities Fibrous proliferation Focal laser scar Nerve fibre layer infarcts Hard exudate Hemorrhage Macular edema Microaneurysm Clinically significant macular edema Panretinal photocoagulation scars Neovascularization of disc Neovascularization elsewhere
Sensitivity, %
Specificity, %
40.0
28.6
45.5 33.3 71.4 70.4 83.8 86.2 82.9 95.3
50.0 100.0 55.6 73.1 70.5 83.3 88.9 94.3
94.1
100.0
100.0 100.0 100.0
100.0 100.0 100.0
relation coefficient 0.86 [p < 0.001]) (Table 4). Severe hemorrhage (SP2A or greater) was identified in two eyes with both techniques. Venous beading greater than SP6A was found in one eye with digital imaging. No eyes were identified with venous beading greater than SP6A with slide film. IRMAs greater than SP8A were identified in one eye with digital imaging and one
neovascularization of the disc (Table 3). The specificity of digital imaging ranged from 28.6% for venous beading to 100.0% for neovascularization of the disc. No venous loops were identified with either slide film or digital imaging. Severe NPDR was identified in three eyes with slide film and four eyes with digital imaging (Pearson's cor-
Table 4-Correlation between stereoscopic digital imaging and slide film in the identification of severe nonproliferative diabetic retinopathy No. of eyes Diabetic retinopathy feature Four fields of severe hemorrhage Two fields of venous beading One field of intraretinal microvascular abnormalities Severe nonproliferative diabetic retinopathy *Two-tailed, n
192
= 241, p < 0.00 I.
CAN J OPHTHALMOL-VOL. 36, NO.4, 2001
Slide film
Digital imaging
Pearson's correlation coefficient*
2
2
1.00 0.00
0
0.00
3
4
0.86
Digital imaging vs. slide film- Tennant et al
different eye with slide film. In both instances the eye identified with IRMAs greater than SP8A with one imaging technique was identified as having IRMAs less than SP8A with the other technique. High-risk proliferative diabetic retinopathy was identified in four eyes with both digital imaging and slide film. Neovascularization of the disc was found in two eyes (less than SPlOA in one eye and greater than SPlOA in one eye). Neovascularization elsewhere was identified in two eyes with both techniques (less than 1/ disc area in one eye and greater than '1 disc area in 2 2 one eye). No eyes were identified with preretinal hemorrhage or vitreous hemorrhage. One eye was identified with three of four high-risk characteristics with both imaging techniques (Pearson's correlation coefficient 1.00 [p < 0.001]). INTERPRETATION
Diabetic retinopathy was present in about half (47.1%) of all patients assessed as part of this pilot study. Although the sample population was not selected randomly, and enrolment bias may have been present, our study does provide an estimate of the current level of diabetic retinopathy in a largely aboriginal population in northern Alberta. In addition, the proportion of patients with treatable disease, 9.9%, is consistent with rates reported in other epidemiologic studies. 13 ·23 Identification of microaneurysms, the cornerstone of early diabetic retinopathy, was similar with slide film and digital imaging, with a correlation coefficient of 0.92. The identification of nonproliferative disorders, including hemorrhage, nerve fibre layer infarcts and hard exudate, also correlated well with slide film. This demonstrates the potential of high-resolution digital imaging to identify early diabetic retinopathy (Fig. 1, top). The correlation coefficient for identifying severe NPDR was 0.86. Correlation was less when identifying IRMAs (0.45) and venous beading (0.32). Following the masked comparison of digital and slide film photographs, discrepancies were identified. Discrepancies were then correlated to individual differences between each digital and film image in an unmasked fashion. The poor correlation between slide film and digital imaging for IRMAs was attributed to two factors. The areas of IRMAs were small and subtle, which led to "missed" identification if image detail was not perfect (Fig. 1, bottom). In addition, the resolution of digital imaging (2008 X 3040 pixels)
was less than with slide film. This level of resolution led to difficulties in identifying IRMAs with certainty. The poor correlation between slide film and digital imaging in the identification of venous beading was attributed to the very low incidence and severity of venous beading in the patients included in the study. In fact, only one eye was found to have venous beading greater than SP6A with either slide film or digital imaging. Lim and associates 9 recently reported the results of a comparison between a nonmydriatic digital camera (resolution 640 X 480 pixels) and slide film. They found low sensitivity for almost all features of diabetic retinopathy, including microaneurysms (46%), dot or blot hemorrhages (71%), flame (50%), hard exudates (64% ), nerve fibre layer infarcts (50%), clinically significant macular edema (30% ), venous beading (50%), neovascularization of the disc (25%) and neovascularization elsewhere in the retina (29% ). The sensitivity of digital imaging in identifying IRMAs (found in one eye) was 100%. Specificity values were much higher than sensitivity values, with many being greater than 90%. We used a digital camera with much greater resolution and therefore had higher sensitivity values for all features except IRMAs and venous beading. In our study identification of neovascularization of the disc (two eyes) and neovascularization elsewhere (two eyes) (Fig. 2, top) was identical with slide film and digital imaging. One patient had three of four high-risk characteristics and was referred to a retinal specialist for treatment. Although the correlation between slide film and digital imaging for severe NPDR and proliferative disease was excellent, only two patients had retinopathy this severe. Interpretation of these findings is therefore limited. Stereoscopic digital imaging had 94.1% sensitivity and 100% specificity in identifying clinically significant macular edema (Fig. 2, bottom). All patients identified with clinically significant macular edema with slide film were referred to a tertiary centre for focal laser treatment. Contact lens biomicroscopy was performed on each eye, confirming clinically significant macular edema in each case. Stereoscopic digital imaging did not identify one eye determined to have clinically significant macular edema on slide film. Following completion of the masked study this image pair was unmasked and reanalysed. It was discovered that although both imaging techniques provided relatively poor detail of the
CAN J OPHTHALMOL-VOL. 36, NO.4, 2001
193
Digital imaging vs. slide film- Tennant et al
Fig. 1-Top left: Fundus photograph obtained with digital camera from patient with diabetic retinopathy, showing microaneurysms. Top right: Magnified view of area of interest. Bottom left: Digital image, showing intraretinal microvascular abnormalities. Bottom right: Magnified view of area of interest.
retina owing to early cataract and poor image focus, the digital image pair was of inferior quality. Even with repeat image analysis clinically significant macular edema could not be identified with any degree of certainty when viewing the digital images. Fortunately, this patient had been identified by digital imaging as having clinically significant macular edema in the fellow eye and would therefore have been referred for treatment by a retinal specialist.
194
CAN
J OPHTHALMOL-VOL. 36, NO. 4, 200 I
CONCLUSION
Stereoscopic digital imaging is both sensitive and specific in identifying diabetic retinopathy as seen on slide film. There is a high degree of correlation between slide film and digital imaging when identifying the presence of rnicroaneurysms, hemorrhage and clinically significant macular edema. Despite the very low incidence of neovascularization of the disc,
Digital imaging vs. slide film- Tennant et al
1-ig. 2-1 op lett: LJigital image, showing neovascularization elsewhere in retina (other than ot disc). I op right: Magnitied view ot area of interest. Bottom left: Digital image, showing clinically significant macular edema. Bottom right: Magnified view of area of interest.
neovascularization elsewhere in the retina and severe NPDR, the correlation between slide film and digital imaging for identifying these features was high. This novel technology may play a role in the screening and diagnosis of treatable diabetic retinopathy via teleophthalmology. This study was supported by the University of Alberta, Edmonton, MacDonald Dettwiler and Associates, Vancouver, Precam and Associates, Ottawa, and the Canadian Space Agency, Ottawa. None of the authors hold any proprietary interests in any of the products mentioned in this article.
REFERENCES
1. Diabetic Retinopathy Study Research Group. Preliminary
report on effects of photocoagulation therapy. Am J Ophthalmo/1976;94:505-37. 2. MacMillan HL, MacMillan AB, Offord DR, Dingle JL. Aboriginal health. Can Med Assoc J 1996;155: 1569-77. 3. Young TK, Shah C. Extent and magnitude of the problem. In: Young TK, editor. Diabetes in the Canadian Native population: bicultural perspectives. Toronto: Canadian Diabetes Association; 1987. p. 11-25. 4. Evers S, McCracken E, Antone T, Deagle G. Prevalence of diabetes in Indians and Caucasians living in southwestern Ontario. Can J Public Health 1987;78:240-3.
CAN J OPHTHALMOL-VOL. 36, NO.4, 200 I
195
Digital imaging vs. slide film- Tennant et al
5. McGregor G. Transmitting digital images of the fundus for the West Virginia Diabetic Eye Care Project. J Ophthalmic Photogr 1994;16:87-98. 6. YoungS, George LD, Lusty J, Owens DR. A new screening tool for diabetic retinopathy: the Canon CR5 45 NM retinal camera with Frost Medical Software RIS-Lite digital imaging system. J Audiov Media Med 1997;20:11-4. 7. Li HK. Telemedicine and ophthalmology. Surv Ophthalmol 1999;44:61-72. 8. LinDY, Blumenkranz MS, Brothers R. The role of digital fundus photography in diabetic retinopathy screening. Diabetes Techno[ Ther 1999;1:477-87. 9. Lim JI, LaBree L, Nichols T, Cardenas I. A comparison of digital nonmydriatic fundus imaging with standard 35millimeter slides for diabetic retinopathy. Ophthalmology 2000;107:866-76. 10. Ryder REJ, Kong N, Bates AS, Sim J, Welch J, Kritzinger EE. Instant electronic imaging systems are superior to Polaroid at detecting sight-threatening diabetic retinopathy. Diabetic Med 1998;15:254--8. 11. Diabetic Retinopathy Study Group. Report 7: a modification of the Airlie House classification of diabetic retinopathy./nvest Ophthalmol Vis Sci 1981;21:210-26. 12. Klein R, Klein BEK, Moss SE, Davis MD, DeMets DL. The Wisconsin Epidemiologic Study of Diabetic Retinopathy. IX. Four-year incidence and progression of diabetic retinopathy when age at diagnosis is less than 30 years. Arch Ophthalmoll989;l07:237-43. 13. Mitchell P, Smith W, Wang JJ, Attebo K. Prevalence of diabetic retinopathy in an older community. The Blue Mountains Eye Study. Ophthalmology 1998;105:406-11. 14. Martin JD, Yidegiligne HM. The cost-effectiveness of a retinal photography screening program for preventing diabetic retinopathy in the First Nations diabetic population in British Columbia, Canada./nt J Circumpolar Health 1998; 57(suppl1):379-82.
196
CAN J OPHTHALMOL-VOL. 36, NO. 4, 200 I
15. Early Treatment Diabetic Retinopathy Study Research Group. Detection of diabetic macular edema: ophthalmoscopy versus photography -Early Treatment Diabetic Retinopathy Study report #5. Ophthalmology 1989;96: 746-51. 16. Early Treatment Diabetic Retinopathy Study Research Group. Grading diabetic retinopathy from stereoscopic color fundus photographs - an extension of the modified Airlie House classification. ETDRS report number 10. Ophthalmology 1991;98(5 suppl):786-806. 17. Wong D. Textbook of ophthalmic photography. New York: Inter-Optics Publications; 1982. p. 76. 18. Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical epidemiology: a basic science for clinical medicine. Toronto: Little, Brown and Company; 1991. p. 69-152. 19. Ferris F. Early photocoagulation in patients with either type I or type II diabetes. Trans Am Ophthalmol Soc 1996;94:505-37. 20. Early Treatment Diabetic Retinopathy Study Research Group. Fundus photographic risk factors for progression of diabetic retinopathy. ETDRS report number 12. Ophthalmology 1991;98:823-33. 21. Diabetic Retinopathy Study Research Group. Four risk factors for severe visual loss in diabetic retinopathy. Arch Ophthalmoll979;97:654-5. 22. Early Treatment Diabetic Retinopathy Study Research Group. Photocoagulation of macular edema. Early Treatment Diabetic Study report #1. Arch Ophthalmol 1985; 103:1796-806. 23. Leske MC, Wu SY, Hyman L, Li X, Hennis A, Connell AM, et al. Diabetic retinopathy in a black population: the Barbados Eye Study. Ophthalmology 1999;106:1893-9.
Key words: diabetic retinopathy, screening, teleophthalmology, imaging techniques, digital stereoscopic photography