Use of offline artificial intelligence in a smartphone-based fundus camera for community screening of diabetic retinopathy

Purpose:The aim of the study was to analyse the reliability of an offline artificial intelligence (AI) algorithm for community screening of diabetic retinopathy.Methods:A total of 1378 patients with diabetes visiting public dispensaries under the administration of the Municipal Corporation of Greater Mumbai between August 2018 and September 2019 were enrolled for the study. Fundus images were captured by non-specialist operators using a smartphone-based camera covering the posterior pole, including the disc and macula, and the nasal and temporal fields. The offline AI algorithm on the smartphone marked the images as referable diabetic retinopathy (RDR) or non-RDR, which were then compared against the grading by two vitreoretinal surgeons to derive upon the sensitivity and specificity of the algorithm.Results:Out of 1378 patients, gradable fundus images were obtained and analysed for 1294 patients. The sensitivity and specificity of diagnosing RDR were 100% (95% CI: 94.72–100.00%) and 89.55% (95% CI: 87.76–91.16%), respectively; the same values for any diabetic retinopathy (DR) were 89.13% (95% CI: 82.71–93.79%) and 94.43% (95% CI: 91.89–94.74%), respectively, with no false-negative results.Conclusion:The robustness of the offline AI algorithm was established in this study making it a reliable tool for community-based DR screening.     

Evaluation of an AI system for the detection of diabetic retinopathy from images captured with a handheld portable fundus camera: the MAILOR AI study

ObjectivesTo evaluate the performance of an artificial intelligence (AI) system (Pegasus, Visulytix Ltd., UK*) at the detection of diabetic retinopathy (DR) from images captured by a handheld portable fundus camera.MethodsA cohort of 6404 patients (~80% with diabetes mellitus) was screened for retinal diseases using a handheld portable fundus camera (Pictor Plus, Volk Optical Inc., USA) at the Mexican Advanced Imaging Laboratory for Ocular Research. The images were graded for DR by specialists according to the Scottish DR grading scheme. The performance of the AI system was evaluated, retrospectively, in assessing referable DR (RDR) and proliferative DR (PDR) and compared with the performance on a publicly available desktop camera benchmark dataset.ResultsFor RDR detection, Pegasus performed with an 89.4% (95% CI: 88.0–90.7) area under the receiver operating characteristic (AUROC) curve for the MAILOR cohort, compared with an AUROC of 98.5% (95% CI: 97.8–99.2) on the benchmark dataset. This difference was statistically significant. Moreover, no statistically significant difference was found in performance for PDR detection with Pegasus achieving an AUROC of 94.3% (95% CI: 91.0–96.9) on the MAILOR cohort and 92.2% (95% CI: 89.4–94.8) on the benchmark dataset.ConclusionsPegasus showed good transferability for the detection of PDR from a curated desktop fundus camera dataset to real-world clinical practice with a handheld portable fundus camera. However, there was a substantial, and statistically significant, decrease in the diagnostic performance for RDR when using the handheld device.Subject terms: Retinal diseases, Physical examination     

Evaluation of a portable fundus camera for use in the teleophthalmologic diagnosis of glaucoma.

PURPOSE: The digital images of the optic disk from a portable fundus camera were evaluated for suitability in teleophthalmologic screening for glaucoma. METHODS: Fifty-one eyes of 27 consecutive patients from our glaucoma clinic were dilated and photographed with a Zeiss FF retinal camera (Carl Zeiss, Oberkochen, Germany) and a portable Nidek NM-100 (Nidek, Tokyo, Japan) fundus camera. Digital images from the portable fundus camera were digitized, compressed and stored in a Fujix DF-10M (Fuji, Tokyo, Japan) digitizer. Lossy compressed digital images and photographs from the Zeiss camera were presented separately in random order to three ophthalmologists for estimation of vertical cup:disk ratios (VCDR) and to evaluate image quality as good, acceptable, or unacceptable for screening glaucoma. Gold standard VCDRs were measured from monoscopic photographic slides obtained using the Zeiss camera by a fourth ophthalmologist. RESULTS: Measurement of agreement (Kappa values) between estimated VCDR of digital images and photographs by the three ophthalmologists were 0.52, 0.38, and 0.50 respectively. Agreement between gold standard and estimated VCDR from photographs were 0.87, 0.45, and 0.84 respectively (specificity between 79% and 97%, sensitivity between 70% and 95%). Kappa values obtained between gold standard and estimated VCDR from digital images were 0.52, 0.49, and 0.49, respectively (specificity between 68% and 79%, sensitivity between 67% and 87%. CONCLUSION: Moderate to good agreement indicates that the digital images from the portable fundus camera may be suitable for optic disk assessment in the current configuration. This easy to use Nidek hand-held camera could be a viable instrument for teleophthalmology if a better digitizing system is incorporated to improve the quality of the images.     

Screening for glaucoma with a non-mydriatic fundus camera

183 first-degree relatives of glaucoma patients were photographed by a technician with non-mydriatic fundus camera in order to study the suitability of wide angle black-and-white fundus photographs in screening for glaucoma. The success rate of photography was 92%. The optic disc and retinal nerve fiber layer abnormalities were evaluated from the photographs by an ophthalmologist. 31 subjects (17%) were referred to further ophthalmological examinations. We found 6 (3%) new glaucomas. In addition, in 6 patients (3%) retinal nerve fiber layer defect was the only abnormality, 5 subjects (3%) showed a hemorrhage and 2 eyes had collateral vessels as a sign of asymptomatic venous stasis change at the optic disc. Only 1 of the 6 (17%) patients with glaucoma would have been found with tonometry alone. The results of this study indicate that non-mydriatic retinal camera is a useful tool in screening for glaucoma.     

Fred Hollows lecture: digital screening for eye disease

The purpose of this study was to explore progress, in the adaptation to community screening for blinding eye disease, of digital imaging devices and technology for storage and transmission. Available imaging systems were compared to gold standard clinical photography in terms of sensitivity and specificity for diagnosis of common blinding eye conditions. Since the use of expensive non-portable imaging devices is likely to be limited for widespread community screening purposes, a portable fundus camera (Nidek, Chiyoda-ku, Japan) and a prototype monocular digital indirect ophthalmoscope constructed at the Lions Eye Institute (LEI) were selected for comparative trials for the screening of optic disc cupping, glaucoma and clinical signs of diabetic retinopathy. Fifty-one eyes of 27 consecutive patients being assessed at the LEI clinic for glaucoma were dilated and photographed with a Zeiss retinal camera, and digital images were taken with the portable Nidek NM100 fundus camera (Carl Zeiss, Oberkochen, Germany) or with a prototype digital monocular indirect ophthalmoscope. Vertical cup: disc ratios (VCDR) were measured on the disc photographs by one ophthalmologist while three other clinicians were presented with compressed digital images in random order to estimate VCDR. Field trials were also carried out to demonstrate the practicality of compression, local storage and then transmission by mobile telephone ISDN lines and satellite, of optic discs and fundus images of patients with diabetes in either rural Western Australia or Surabaya, Indonesia. Kappa values of correlations of measurement of agreement between measured and estimated VCDR were 0.87, 0.45 and 0.84, respectively, for the three observers, corresponding to a specificity of 79-97% and a sensitivity of 70-95%. The portable Nidek fundus camera was also assessed for specificity and sensitivity in the diagnosis of diabetic retinopathy in comparison to standard Zeiss fundus camera photographs. Of 49 eyes in 25 consecutive patients attending the LEI clinic for assessment of diabetic retinopathy, three ophthalmologists assessed photographs and images in random order. When used for screening diabetic retinopathy, the digital images of the Nidek camera were graded as adequate quality in only 56% of eyes compared to 93% of the photographs. The kappa value of agreement in analysis of diabetic retinopathy was only 0.30. The prototype digital monocular indirect ophthalmoscope compared favourably with the Nidek camera. At 1:5 compression, images of size 36 kB transmitted from Surabaya to Perth took 29 s on the mobile telephone, while uncompressed images took 170 s. Images compressed 1:5 were transmitted in 60 s using the satellite telephone, while the uncompressed images took 240 s. Satellite transmission was more expensive but the lines were more stable than telephone connections from Indonesia. Digital imaging is becoming a powerful tool for ophthalmology in clinical records, teaching and research, and interoffice diagnostic opinions. It also has enormous potential for community screening for blinding eye diseases, such as glaucoma and diabetic retinopathy. Inexpensive portable imaging devices that are easy to use, and on which local health workers might be trained, must be developed and validated in terms of sensitivity and specificity of performance. The technology of image capture, image compression, transmission, data base storage and analysis is rapidly evolving and becoming less expensive.     

Utility of 1% Tropicamide in Improving the Quality of Images for Tele-Screening of Diabetic Retinopathy in Patients with Dark Irides

Purpose: To compare the quality of fundus photographs taken before and after instillation of one drop of tropicamide.

Methods: The 45º fundus photographs were taken with a non-mydriatic fundus camera in three conditions of the pupil; pre-mydriatic, 10 minutes after one drop of tropicamide, and fully dilated. Two photographs were taken in each condition; one centered on the macula and the other on the optic disc. Two vitreoretinal specialists graded the images.

Results: A total of 1768 fundus photographs of 149 diabetic patients with dark irides were included. There were more ungradable images (38.1% and 50.3%, graders 1 and 2, respectively) in the non-mydriatic state than partially- (4.6% and 11.5%) or fully-dilated (15.4% and 10.0%) conditions (p < 0.001, both graders). Partially and fully dilated states had similar rates of ungradable images (p = 0.56 and p = 0.54, graders 1 and 2, respectively). Test–retest reliability (repeatability) was 92.5% and 74.3% for the two graders, respectively. Inter-grader agreement was moderate (Kappa = 0.50).

Conclusion: Non-mydriatic fundus photographs have a high rate of ungradable images in patients with dark irides. Instillation of only one drop of tropicamide improves the quality of fundus photographs, which is not furthered by adding more drops. This strategy can be used in tele-ophthalmology programs.     

免散瞳眼底照相技术在学龄前儿童眼底病筛查中的应用

目的：观察免散瞳眼底照相机在检查中所见的学龄前儿童眼底病变的发病情况,并与直接检影镜检查结果进行对比分析。方法：收集2012-04/2013-10来我院眼科门诊就诊的学龄前儿童3 896例7 760眼,采用日本 Topcon TRC-NW300彩色荧光眼底照相机拍照,图像及时保存,并进行直接检影镜检查。结果：免散瞳眼底照相技术的检出率较直接检影镜高。3 896例7 760眼患儿中,检出41眼（1.05%）眼底异常者。其中视网膜有髓鞘神经纤维（24.39%）、牵牛花综合征（21.95%）、视网膜色素变性（14.63%）、先天性视网膜劈裂（12.20%）; 而儿童眼病非单一发生,常伴有视力异常（68.30%）、屈光异常（63.41%）、斜视（19.51%）。结论：免散瞳眼底照相技术具有不用药物散瞳,学龄前儿童易于接受,图像结果即刻显示,能够直观、清晰显示眼底各种病变,对学龄前儿童眼病筛查有重要意义。     

Effect of mydriasis and different field strategies on digital image screening of diabetic eye disease

AIMS: To assess the effects of (1) mydriasis and (2) single versus three field photography on screening for diabetic eye disease using digital photography METHOD: Slit lamp examination findings were compared to digital fundal photographs for the detection of any retinopathy and for referable retinopathy in 398 patients (794 eyes). A Topcon TRC-NW6S digital non-mydriatic fundus camera was used. Three photographic strategies were used: undilated single field, dilated single field, and dilated multiple fields. The photographs were presented in random order to one of two retinal screeners. For the single field photographs the screeners were masked to the use of mydriatics. In 13% of fundal photographs, grading was performed by both, rather than just one grader. RESULTS: Mydriasis reduced the proportion of ungradable photographs from 26% to 5% (p<0.001). Neither mydriasis nor three field photography improved the sensitivity or specificity for the detection of any retinopathy or of referable retinopathy when compared with undilated single field photography. The sensitivity and specificity for detecting referable retinopathy using undilated single field photography was 77% (95% CI 71 to 84) and 95 % (95% CI 93 to 97) respectively. Using dilated single field photography the figures were 81% (95% CI 76 to 87) and 92% (95% CI 90 to 94) respectively. Using dilated three field photography the figures were 83% (95% CI 78 to 88) and 93% (95% CI 91 to 96) respectively. Intergrader reliability for the detection of referable retinopathy in gradable photographs was excellent (Kappa values 0.86-1.00). CONCLUSIONS: Mydriasis reduces the technical failure rate. Mydriasis and the three field photography as used in this study do not increase the sensitivity or specificity of detecting diabetic retinopathy.     

The follow-up of patients screened for glaucoma with non-mydriatic fundus photography

In order to evaluate the value of photographic screening in predicting progressive glaucomatous damage, we re-examined 26 subjects 5 years after the initial screening. Of the 26 patients 16 had typical glaucomatous optic disc and visual field abnormalities (n=7), retinal nerve layer damage (n=6), or other risk factors of glaucoma (n=3). In 10 of 26 patients suspected of having glaucoma, no abnormalities were initially confirmed. Of the 16 eyes with initially abnormal findings, 10 (63%) showed progressive changes during the 5-year follow-up period. The 10 initially suspected cases have remained healthy throughout the follow-up, giving a false positive rate of 5.5%. The results of this study indicate that it is possible to identify correctly patients with progressive glaucomatous changes with a non-mydriatic fundus camera.This study was presented in part at the ARVO Meeting May 3–8, 1992.     

A comparison of digital nonmydriatic fundus imaging with standard 35-millimeter slides for diabetic retinopathy

PURPOSE: The purpose of this study was to compare nonmydriatic digital images with 35-mm slide images for the detection of diabetic retinal findings. STUDY DESIGN: Comparative, observational case series. PARTICIPANTS: Twenty-two patients with diabetes. METHODS: Diabetic patients underwent digital (nonmydriatic fundus camera attached to a digital back) and standard dilated 35-mm retinal photography of three areas: posterior pole, nasal retina, and temporal retina. The images were reviewed by a single masked grader for the presence or absence of specific retinal findings. PRIMARY OUTCOME MEASURES: Presence or absence of neovascularization of the disc (NVD), neovascularization elsewhere (NVE), venous beading (VB), nerve fiber layer (NFL) hemorrhage, dot-blot hemorrhage, microaneurysm (MA), clinically significant macular edema, cotton wool spot, intraretinal microvascular anomaly (IRMA), hard exudate (HE), and retinal pigment epithelial (RPE) pigmentary changes. RESULTS: Forty eyes of 22 patients underwent both imaging procedures. The agreement between image type was highest for IRMA (97.5%) and VB (95%) and lowest for RPE pigmentary changes (65%) and MA (62.5%). Sensitivity ranged from 25% (NVD) to 100% (VB). Specificity ranged from 90% (RPE pigmentary changes) to 100% (NVD, NVE, VB, NFL hemorrhage, HE). Positive predictive value ranged from 50% (IRMA) to 100% (NVD, NVE, HE, NFL, VB). Negative predictive value ranged from 48% (MA) to 100% (IRMA). CONCLUSIONS: Nonmydriatic digital fundus imaging for detection of diabetic retionopathy has a low sensitivity rate and a high specificity rate and is less clinically useful than standard dilated 35-mm fundus slide images.     

A comparison of digital retinal image quality among photographers with different levels of training using a non-mydriatic fundus camera

PURPOSE: To evaluate the quality of digital retinal images taken by three photographers with different levels of photographic training, using a non-mydriatic fundus camera. METHODS: This study compares 45-degree digital retinal images taken with a non-mydriatic fundus camera by three different photographers with different levels of photographic training: (I) A professional ophthalmic photographer with 20 years of experience; (2) a non-professional photographer with 2 days of photographic training and experience with 50 patients; (3) a non-professional photographer with 1 hour of photographic training and experience with 10 patients. The quality of the photographs was evaluated by the consensus of two retina specialists. RESULTS: Sixty-four (64) eyes of 33 subjects were imaged by the three photographers for a total of 192 images. Thirty-four eyes were photographed in the non-dilated state. The trained ophthalmic photographer and the two non-professional photographers did not have statistically significant differences in image quality based on the image evaluations. (Chi-square P-value: 0.57). This finding was consistent for eyes in both the non-dilated and dilated state. CONCLUSIONS: Fundus image quality for images taken with a non-mydriatic camera were not significantly different among three photographers with different levels of training.     

Diabetic retinopathy screening using digital non-mydriatic fundus photography and automated image analysis

PURPOSE: To investigate the use of automated image analysis for the detection of diabetic retinopathy (DR) in fundus photographs captured with and without pharmacological pupil dilation using a digital non-mydriatic camera. METHODS: A total of 83 patients (165 eyes) with type 1 or type 2 diabetes, representing the full spectrum of DR, were photographed with and without pharmacological pupil dilation using a digital non-mydriatic camera. Two sets of five overlapping, non-stereoscopic, 45-degree field images of each eye were obtained. All images were graded in a masked fashion by two readers according to ETDRS standards and disagreements were settled by an independent adjudicator. Automated detection of red lesions as well as image quality control was made: detection of a single red lesion or insufficient image quality was categorized as possible DR. RESULTS: At patient level, the automated red lesion detection and image quality control combined demonstrated a sensitivity of 89.9% and specificity of 85.7% in detecting DR when used on images captured without pupil dilation, and a sensitivity of 97.0% and specificity of 75.0% when used on images captured with pupil dilation. For moderate non-proliferative or more severe DR the sensitivity was 100% for images captured both with and without pupil dilation. CONCLUSION: Our results demonstrate that the described automated image analysis system, which detects the presence or absence of DR, can be used as a first-step screening tool in DR screening with considerable effectiveness.     

Screening for glaucoma in a general population with the non-mydriatic fundus camera and the frequency doubling perimeter

PURPOSE: To evaluate the usefulness of non-mydriatic fundus camera (NMFu-camera) and frequency doubling perimeter (FDP) for detecting glaucoma in a general population. METHODS: This prospective observational multicenter study consisted in screening for glaucoma in the populations of three Belgian cities. Intraocular pressure (IOP) was measured with non-contact pneumo-tonometer (NCT) and applanation tonometry (AT) if NCT IOP was > or = 17 mmHg. Visual field was screened with FDP (C-20-5) and digitized optic disc photographs (ODPs) were taken with NMFu-camera. FDP was considered abnormal if at least one defective point was found. ODPs were graded as normal or glaucomatous by consensus of three glaucoma specialists. Optic disc and visual field results were matched per eye. Subjects with known ocular hypertension and/or treated primary open angle glaucoma were excluded from the analysis. RESULTS: A total of 1620 subjects were included in the study. Their mean age was 63.2 years. AT IOP was > 21 mmHg in 8.2%. A total of 98.1% of ODPs could be interpreted. Glaucomatous optic discs were detected in 3.5% of the subjects. In this group only 24% had an AT IOP > or = 22 mmHg. FDP was abnormal in 44.5%. The sensitivity and specificity of FDP to identify patients with an optic disc graded as glaucomatous were 58.6% and 64.3% respectively. CONCLUSIONS: The combined use of the NMFu-camera and the FDP is a feasible method for an initial glaucoma mass screening. NMFu-camera may be a useful and quick method to screen for glaucomatous damage in a community. FDP in screening strategy was revealed to be not sensitive enough when setting the cut-off value at one defective test location. IOP measurements were confirmed to be a poor tool to detect glaucomatous damage.     

Semiautomated quantification of β‐zone parapapillary atrophy using blue light fundus autofluorescence

Purpose: To evaluate the reproducibility of measurements of area of β‐zone parapapillary atrophy (β‐PPA) using blue laser fundus autofluorescence (FAF) and confocal scanning laser ophthalmoscopy reflectance (CSLO) measurements and to assess agreement between the two imaging modalities. Methods: Sixty‐five eyes of 45 patients (mean age, 68.2 ± 11.3 years) with established or suspected glaucoma from the Diagnostic Innovations in Glaucoma Study (DIGS) were prospectively included. FAF scans were obtained with the Spectralis HRA+OCT and CSLO reflectance images with the HRTII (both from Heidelberg Engineering, Heidelberg, Germany). Two masked graders independently measured β‐PPA area on 3 consecutive scans using the semi‐automated BluePeak RegionFinder software (BPRF) and on CSLO reflectance images using the optic disc contour line. Reproducibility of β‐PPA area measurements was assessed using intraclass correlation coefficients (ICC). Results: Intragrader reproducibility was 0.997 (95% CI, 0.996–0.998) and 0.995 (95% CI, 0.992–0.996) for grader 1 and 2, respectively, using FAF‐BPRF, and by CSLO, it was 0.991 (95% CI, 0.986–0.994) and 0.988 (95% CI, 0.982–0.992). Intergrader agreement (ICC) was 0.53 (95% CI, 0.331–0.685) for FAF‐BPRF and 0.404 (95% CI, 0.149–0.601) for CSLO (comparison between ICC, p = 0.368). Agreement (ICC) between the two devices was worse for grader 1 (0.356; 95% CI, 0.129–0.549) than grader 2 (0.856; 95% CI, 0.774–0.910) (p < 0.001). Conclusions: Despite excellent intragrader reproducibility for β‐PPA measurements with FAF‐BPRF and CSLO, intergrader reproducibility is low to moderate. Measurements of β‐PPA area obtained with the two instruments are of moderate agreement and, therefore, are not interchangeable.     

单视野免散瞳数码眼底照相筛查糖尿病视网膜病变的敏感性和特异性

目的:单视野免散瞳眼底照相和散瞳直接眼底镜检查与荧光素眼底血管造影(fundus fluorescein angiography,FFA)相比较,评价其筛查糖尿病视网膜病变(diabetic retinopathy,DR)的敏感性和特异性。方法:1型或2型糖尿病患者93例186眼,先后进行单视野免散瞳数码眼底照相、散瞳直接眼底镜检查和FFA,以FFA诊断结果作为比较标准,评价单视野免散瞳数码眼底照相和散瞳直接眼底镜筛查DR的敏感性和特异性。结果:单视野免散瞳数码眼底照相检出DR的敏感性和特异性分别为80.4%和94.7%,而散瞳直接眼底镜检出DR的敏感性和特异性分别为64.2%和84.2%。当把筛查阈值下调至中度非增殖性DR(M-NPDR)后,单视野免散瞳数码眼底照相检出DR的敏感性和特异性分别提高为88.9%和98.4%,散瞳直接眼底镜检出DR的敏感性和特异性也均有提高,分别为71.5%和96.7%。结论:单视野免散瞳数码眼底照相是筛查DR的有效工具。     

Sensitivity and specificity of handheld fundus cameras for eye disease: A systematic review and pooled analysis

In order to evaluate the accuracy of commercially available handheld fundus cameras for a variety of ophthalmic diagnoses, we conducted a systematic review, searching PubMed and PubMed Central and performing a bivariate analysis to determine the pooled sensitivity and specificity of handheld fundus cameras. Eleven studies validating handheld fundus cameras against a gold-standard method for disease diagnosis were included. For nonmydriatic images, pooled sensitivity was 83% (95% confidence interval (CI): 77-88%) and specificity was 92% (95% CI: 79-97%). For mydriatic images, pooled sensitivity was 87% (95% CI: 79-92%) and specificity was 90% (95% CI: 78-96%). Overall pooled sensitivity was 85% (95% CI: 80-89%) and specificity was 91% (95% CI: 83-95%). Of the 11 studies included, 5 assessed the diagnosis of diabetic retinopathy, for which sensitivity was 87% (95% CI: 80-92%) and specificity was 95% (95% CI: 85-98%). For all other diagnoses combined, sensitivity was 81% (95% CI: 74-87%) and specificity was 83% (95% CI: 76-89%). These findings suggest that handheld fundus cameras are capable of achieving acceptable sensitivity and specificity values for eye disease, with mydriatic images being more sensitive for disease. Diabetic retinopathy was the single diagnosis with the strongest data to support the use of handheld fundus cameras for disease screening.     

Screening for diabetic retinopathy by one-field, non-mydriatic, 45 degrees digital photography is inadequate

PURPOSE: To evaluate the sensitivity and specificity of one-field, non-mydriatic, 45 degrees digital photography for screening for diabetic retinopathy compared to indirect ophthalmoscopy using a slit-lamp, the reference standard. METHODS: A total of 100 consecutive diabetic patients (200 eyes) who underwent digital fundus photography and ocular examinations from June 2002 to November 2002 were included in this retrospective study. The patients, recruited from a hospital-based, retina referral practice, underwent 45 degrees, non-mydriatic, digital fundus photography using a non-mydriatic fundus camera. One image was obtained focusing the mid fundus between the optic disc and the macula. The fundus images were printed and graded by endocrinologists and a retinal specialist separately. The patients also underwent complete standard ocular examinations as the reference method for determining diabetic retinopathy, including dilation of their pupils and slit-lamp biomicroscopy done by ophthalmologists. The sensitivity and specificity of the digital photographic method were calculated by comparison to the reference method. RESULTS: The sensitivity and specificity of the retinal specialist's diabetic retinopathy grades were 53.8 and 89.0%, respectively. The sensitivity and specificity of the endocrinologists' grades were 45 and 75.3%, respectively. The false negative rates were 22 and 21.5% for endocrinologists and the retinal specialist, respectively. CONCLUSIONS: Screening for diabetic retinopathy using one-field, non-mydriatic, 45 degrees digital photography is inadequate.     

欧堡全景200Tx激光扫描检眼镜在白内障术后患者眼底检查中的应用

目的：研究欧堡全景200Tx激光扫描检眼镜(简称Optomap 200Tx)在白内障术后患者早期眼底筛查的应用价值。

方法：对2013-11-18/12-31来我院就诊的白内障患者146例161眼,均为连续病例,于术后1wk行非散瞳下直接检眼镜检查、非散瞳下Optomap 200Tx检查、散瞳后裂隙灯下前置镜检查,并对三种检查方法进行对比。

结果：患者146例161眼行非散瞳下直接检眼镜检查、非散瞳下Optomap 200Tx检查、散瞳后前置镜检查分别发现眼底病变40眼(24.8%),59眼(36.7%),61眼(37.9%),需要及时治疗病例9眼(5.6%)。非散瞳下Optomap 200Tx检查结果与散瞳后前置镜检查结果相似,差异无统计学意义,优于非散瞳下直接检眼镜检查,差异有统计学意义(P<0.05)。

结论：白内障患者术前屈光间质混浊,无法详查眼底,故应常规对白内障术后患者行全面眼底筛查。Optomap 200Tx是白内障术后早期眼底筛查的一项有效、便捷的检查手段,与散瞳后裂隙灯下前置镜检查有相似检出率。     

A new handheld fundus camera combined with visual artificial intelligence facilitates diabetic retinopathy screening

AIM: To explore the performance in diabetic retinopathy (DR) screening of artificial intelligence (AI) system by evaluating the image quality of a handheld Optomed Aurora fundus camera in comparison to traditional tabletop fundus cameras and the diagnostic accuracy of DR of the two modalities. METHODS: Overall, 630 eyes were included from three centers and screened by a handheld camera (Aurora, Optomed, Oulu, Finland) and a table-top camera. Image quality was graded by three masked and experienced ophthalmologists. The diagnostic accuracy of the handheld camera and AI system was evaluated in assessing DR lesions and referable DR. RESULTS: Under nonmydriasis status, the handheld fundus camera had better image quality in centration, clarity, and visible range (1.47, 1.48, and 1.40) than conventional tabletop cameras (1.30, 1.28, and 1.18; P<0.001). Detection of retinal hemorrhage, hard exudation, and macular edema were comparable between the two modalities, in principle, with the area under the curve of the handheld fundus camera slightly lower. The sensitivity and specificity for the detection of referable DR with the handheld camera were 82.1% (95%CI: 72.1%-92.2%) and 97.4% (95%CI: 95.4%-99.5%), respectively. The performance of AI detection of DR using the Phoebus Algorithm was satisfactory; however, Phoebus showed a high sensitivity (88.2%, 95%CI: 79.4%-97.1%) and low specificity (40.7%, 95%CI: 34.1%-47.2%) when detecting referable DR. CONCLUSION: The handheld Aurora fundus camera combined with autonomous AI system is well-suited in DR screening without mydriasis because of its high sensitivity of DR detection as well as its image quality, but its specificity needs to be improved with better modeling of the data. Use of this new system is safe and effective in the detection of referable DR in real world practice.     

Evaluation of the effectiveness of ophthalmic assistants as screeners for glaucoma in North India

Aim

To assess whether ophthalmic assistants are effective in screening people for glaucoma in India.

Methodology

The study subjects were examined by both trained ophthalmic assistants and an ophthalmologist in both hospital and community settings. Specific tests for the diagnosis of glaucoma suspects included visual field examination using frequency doubling technology perimetry, intraocular pressure measurement (Tonopen), A-scan central anterior chamber depth measurement and dilated optic disc examination. The findings recorded by the ophthalmic assistants were masked to the ophthalmologist to avoid measurement bias.

Results

In the hospital setting, there was a substantial level of agreement between the ophthalmic assistants and the ophthalmologist in the diagnosis of glaucoma suspects (89.29%, k=0.7, 95% confidence interval (CI)=0.54–0.86). The diagnostic accuracy of the ophthalmic assistants in detecting glaucoma suspects was high for sensitivity (95.2%, 95% CI=91.4–97.7%) but lower for specificity at 71.4% (95% CI=60.0–78.7%).In the community setting, there was a moderate level of agreement between the ophthalmic assistants and the ophthalmologist in the diagnosis of glaucoma suspects (78.23%, k=0.50, 95% CI=0.37–0.64). The diagnostic accuracy of the ophthalmic assistants in detecting glaucoma suspects was moderate for sensitivity (82.9, 95% CI=69.7–91.5%) but lower for specificity at 76.8% (95% CI=72.7–79.5%).

Conclusion

Ophthalmic assistants can be used for opportunistic case detection of glaucoma suspects in the community. Structured training of the ophthalmic assistants together with enhanced clinical experience would improve their performance in detecting glaucoma suspects in the community.