Research

Hemoglobin measurement in the optic nerve head using the Laguna ONhE program

European Glaucoma Society Congress 2018
DOI:
Marta Gonzalez-Hernandez, Silvia Alayon Miranda, Jose Sigut Saavedra, Manuel Gonzalez de la Rosa

Purpose
The Laguna ONhE program divides conventional color images of the optic nerve head (ONH) into 24 sectors using two ellipses, approximately parallel to its edge, and four diametrical lines. It notes the differences between its red and green components and compensates the diversities of spectral composition of the illumination light, the absorption of the lens and the spectral response of the detector used by means a relative measure: the values of the tissue are divided by those obtained in the central vessels.

Previous experience in 700 normal and 494 glaucoma images obtained with Nidek, Kowa and Topcon fundus cameras was used to optimize the Laguna ONhE index “Glaucoma Discriminant Function, GDF“.


Methods
96 healthy subjects and 82 confirmed and suspect glaucoma were examined twice with the Laguna ONhE method (INSOFT, Spain), using images obtained with the Horus DEC- 200 portable fundus camera (MiiS, Taiwan), and once with the Spectralis OCT (Heidelberg, Germany). The images were divided into two groups of better and worst contrast, comparing vessels Vs tissue, using the red and green channels in the optic nerve image. Differences between outcomes were analyzed with the MedCalc 17.9.7 program. 

Results
Both series respectively had contrasts of 1.58±0.33 and 1.95±0.60 (p<0.0001). The Pearson correlation coefficient between GDF and BMO-MRW was 0.827-0.831 in the two groups of images (p<0.0001, Fig 1), between GDF and RNFLT was 0.763-0.766 (p<0.0001) and between BMO-MRW and RNFLT was 0.848 (p<0.0001, Fig 2). Intra-class correlation coefficient between the GDF values of the two exams was 0.957. Using ROC analysis, we calculated the confidence intervals (5%-95%) of the area under the curve, the specificity closest to 95%, and the corresponding sensitivity.


Conclusion
Using a simple manual fundus camera to study the distribution of hemoglobin in the optic nerve achieves a diagnostic capacity of glaucoma almost equivalent, or minimally different, to an OCT, even using images of sub-optimal quality.

Read publication

Reproducibility of the Laguna ONhE application with automatic segmentation of the limits of the optical nerve head, excavation and neuroretinal rim: Validation of new algorithms (Spanish)

DOI:
PhD Student: Marta González Hernández, Supervisor: Prof. José Francisco Sigut Saavedra

Purpose
To compare the diagnostic accuracy and reproducibility of the Laguna ONhE program (with automatic segmentation of the optic nerve head cupping and neuro-retinal rim using new algorithms), with morphological (OCT) and functional (visual field) information.

Methods
96 healthy subjects and 82 glaucomas were examined twice with photographic images obtained with a fundus camera Horus Scope DEC-200 (MiiS) and analyzed with the modified Laguna ONhE (Insoft) software, twice with Cirrus OCT (Zeiss), once with Spectralis OCT (Heidelberg) and once with Octopus 300 TOP-32 (Haag-Streit). Statistics used: Receiver operator characteristic (ROC) curve analysis, Pearson correlations, intra-class correlation coefficients (ICC) and respective confidence intervals (CI), and kappa concordance index.

Results
Laguna ONhE glaucoma discriminant function (GDF) was among the indices of greatest area under the ROC curve (AUROC) (Confidence interval CI=0.87-0.95 in the first examination and 0.86-0.94 in the second), similar to that obtained with Bruch’s membrane opening-minimum rim width (BMO-MRW) of Spectralis (CI=0.91-0.97). Diagnostic concordance between the two was good (kappa =0.639) and similar to that observed, for example, between the set of Spectralis and Cirrus indices (kappa=0.592). Based on hemoglobin information, Laguna ONhE estimation of rim and cup shape and size showed AUROC equivalent to those of Cirrus (Cirrus vertical C/D ratio CI=0.86-0.94, Estimated Laguna ONhE vertical C/D ratio CI=0.83-0.92). For a specificity of 95%, the cut of diagnostic value of the (estimated) Laguna ONhE vertical C/D ratio was 0.59-0.62 and for the (average) Cirrus 0.68-0.75. The reproducibility of Laguna ONhE indices measured with ICC was: GDF (CI=0.88-0.93) and estimated Hb C/D vertical ratio (CI=0.90-0.94). This proved similar to the C/D vertical ratio Cirrus reproducibility (CI=0.95-0.97). Cirrus RNFT reproducibility was slightly better (CI=0.98-0.99). Perimetric indices showed slightly lower diagnostic capacity, but this was not statistically significant with most of others: Mean defect AUROC (CI=0.82-0.91) and square root of loss variance AUROC (CI=0.79-0.89).

Conclusions
Laguna ONhE showed high diagnostic capacity and reproducibility, equivalent to other methods such as OCT. This procedure provides information different from functional or morphological data, related with optic nerve head perfusion. Morphological estimation using Laguna ONhE showed a similar range of diagnostic capacity in the sample analyzed to that measured by OCT.

Read publication

Correlating cup-to-disc ratios measured by HRT-III, SD-OCT and the new color imaging Laguna ONhE procedure (Spanish)

Archives of the Spanish Society of Ophthalmology
I. Rodríguez Una, C.D. Méndez Hernández, F. Sáenz-Francés and J. García Feijóo

Purpose
To examine correlations between cup-to-disc (C/D) ratios determined by the new Laguna ONhE (optic nerve hemoglobin) color imaging procedure, spectral domain optical coherence tomography (OCT), confocal scanning laser tomography using Heidelberg retina tomography (HRT), and examining retinal images.

Methods
C/D ratio measurements were made on 154 eyes of 154 subjects (52 healthy controls, 36 with ocular hypertension and 66 with primary open-angle glaucoma) using the Laguna ONhE, HRT-III (Heidelberg Engineering) and OCT Spectralis (Heidelberg Engineering) instruments and photographs of the optic disc were examined by a blinded observer (experienced glaucoma specialist).

Results
Global intraclass correlation coefficients (ICC) were: 0.379 (95% CI: 0.233–0.508) for Laguna ONhE-HRT, 0.621 (95% CI: 0.513–0.709) for Laguna ONhE-OCT, and 0.558 (95% CI: 0.398–0.678) for the Laguna ONhE-observer, indicating significant agreement in each case (p < 0.001). The highest ICC was recorded for OCT-observer (0.715; 95% CI: 0.605–0.795).

Conclusions
C/D ratios measured using the Laguna ONhE procedure correlated well with OCT measurements and retinography measurements made by an experienced observer. Best correlation was observed for OCT versus observer measurements. Agreement was good between the Laguna ONhE, OCT and observer measurements, and was somewhat lower between HRT and the remaining procedures.

Read publication

Measuring Hemoglobin Levels in the Optic Nerve Head for Glaucoma Management

Glaucoma Imaging. Springer
Manuel Gonzalez-de-la-Rosa, Marta Gonzalez-Hernandez, Carmen Mendez- Hernandez, Elena Garcia-Martin, Francisco Fumero-Bautista, Silvia Alayon, and Jose Sigut

Description of the method Laguna ONhE (used by the RetinaLyze Glaucoma algorithm) that measures the amount of hemoglobin at the optic nerve head from color fundus pictures. It shows the basis of the method and its application to glaucoma and other pathologies that affect the optic nerve.

Read publication

Relationship between Retinal Nerve Fiber Layer Thickness and Hemoglobin Present in the Optic Nerve Head in Glaucoma

Journal of Ophthalmology
Marta Gonzalez-Hernandez, Jose Sigut Saavedra, and Manuel Gonzalez de la Rosa

Purpose
To observe the relationship between topographic hemoglobin levels in the optic nerve head (ONH), the rim thickness (BMO-MRW), and retinal nerve fiber layer (RNFL) thickness.

Methods
96 normal eyes and 82 glaucomas were examined using TOP strategy (Octopus 300 perimeter), SPECTRALIS OCT, and Laguna ONhE program which estimates hemoglobin from conventional color photographs (Horus Scope DEC 200 fundus camera). 

Results
The correlation between Laguna ONhE glaucoma discriminant function (GDF) and SPECTRALIS BMO-MRW was  (), similar to that between the BMO-MRW and BMO-RNFL thicknesses (, ) ( between both  values). GDF correlated well with RNFL thicknesses in the 360 degrees around the nerve, similar to mean perimetric sensitivity (MS) and BMO-MRW. The amount of hemoglobin in the nasal and temporal sectors showed low correlation with superior and inferior RNFL thicknesses. The superotemporal and inferotemporal sectors located on the vertical diameter of the disk showed good intercorrelation but without a clear RNFL topographic relationship. 

Conclusion
GDF showed high correlation with RNFL thickness. Except in the nasal and temporal sectors, ONH hemoglobin correlated well with RNFL thickness.

Read publication

Reproducibility of Optic Nerve Head Hemoglobin Measures

Glaucoma Journal
Carmen Mendez-Hernandez, Julian Garcia-Feijoo, Paula Arribas-Pardo, Federico Saenz-Frances, Ignacio Rodriguez-Una, Cristina Fernandez-Perez and Manuel Gonzalez de la Rosa

Purpose
To evaluate intraobserver, interobserver, within-session and between-session reproducibility of the measurement of optic nerve head (ONH) hemoglobin levels by color analysis using Laguna ONhE [optic nerve hemoglobin (ONH Hb)] program.

Methods
This was an observational prospective study of 29 eyes (11 glaucomatous; 18 healthy eyes). Two examiners obtained 2 retinal photographs (Canon non-mydriatic retinal camera CD-DGi, Canon Inc.,Tokyo, Japan) in 2 testing sessions 3 weeks apart and analyzed the images using Laguna ONhE. The following parameters were quantified: ONH hemoglobin amounts across the whole disc (ONH Hb) and in the vertical disc diameter (8&20 Hb), cup-disc ratio (C/D), and the Glaucoma Discriminant Function (GDF). Agreement was illustrated using the Bland-Altman plots and reproducibility was assessed comparing the intraclass correlation coefficients (ICC).

Results
In session 1, examiner 1 found mean levels of ONH hemoglobin of 67.94±8.70% in healthy eyes and of 57.90±5.36% in glaucomatous eyes. Corresponding values for examiner 2 were 68.27±8.52% and 57.83±4.88%, respectively. ONH Hb and 8&20 Hb measurements were lower in glaucomatous eyes (P=0.002 and P=0.001 respectively). GDF was also more pathologic in glaucomatous group. C/D ratio estimation was greater in the glaucoma group (P=0.003). ONH Hb and 8&20 Hb showed the highest ICCs (all above 0.9). Variability was greater for GDF (ICC>0.8) and C/D ratio estimation (ICC>0.71).

Conclusions
Measurement of ONH Hb levels using the Laguna ONhE program shows high reproducibility both in glaucomatous and nonglaucomatous ONHs.

Read publication

Estimation of the Relative Amount of Hemoglobin in the Cup and Neuroretinal Rim Using Stereoscopic Color Fundus Images

Investigative Ophthalmology & Visual Science
Cristina Pena-Betancor, Marta Gonzalez-Hernandez, Francisco Fumero-Batista, Jose Sigut, Erica Medina-Mesa, Silvia Alayon, and Manuel Gonzalez de la Rosa

Purpose
To calculate the relative amount of hemoglobin (Hb) in sectors of the optic nerve head (ONH) from stereoscopic color fundus images using the Laguna ONhE method and compare the results with the visual field evaluation and optical coherence tomography (OCT).

Methods
Healthy eyes (n = 87) and glaucoma eyes (n = 71) underwent reliable Oculus Spark perimetry and Cirrus OCT. Optical nerve head color images were acquired with a nonmydriatic stereoscopic Wx Kowa fundus camera. Laguna ONhE program was applied to these images to calculate the relative Hb amount in the cup and six sectors of the rim. Receiver operating characteristic (ROC) analysis and correlations between parameters were calculated.

Results
We did not observe any variations in the relative amount of Hb in relation to age in healthy subjects (R(2) = 0.033, P > 0.05). Maximum ROC area confidence intervals were observed for a combination between perimetric indices and the Laguna ONhE Glaucoma discriminant function (0.970-0.899) followed by rim area (0.960-0.883), and mean deviation (MD; 0.944-0.857). In glaucoma cases, relative Hb amount presented significant reduction in all rim sectors, especially 231° to 270° and 81° to 120° (P < 0.001), except in the temporal 311° to 40° (P = 0.11). Perimetry mean sensitivity by sectors was better correlated with respective Hb levels than with rim areas or the corresponding nerve fiber thickness, especially the superior and inferior sectors (P < 0.05).

Conclusions
Visual field sensitivity was better correlated with Hb levels than with rim sector areas or the corresponding nerve fiber thickness. In many cases the remaining rim show low perfusion, especially in the superior and inferior sectors.

Read publication

Measuring Hemoglobin Levels in the Optic Nerve Head: Comparisons with Other Structural and Functional Parameters of Glaucoma

Investigative Ophthalmology & Visual Science
Manuel Gonzalez de la Rosa, Marta Gonzalez-Hernandez, Jose Sigut, Silvia Alayon, Nathan Radcliffe, Carmen Mendez-Hernandez, Julian Garcıa-Feijoo, Isabel Fuertes-Lazaro, Susana Perez-Olivan and Antonio Ferreras

Purpose
We evaluated and compared the ability of a new method for measuring hemoglobin (Hb) levels at the optic nerve head (ONH) to that of visual field evaluation, scanning laser ophthalmoscopy (HRT), scanning laser polarimetry (GDx), and optical coherence tomography (OCT) for diagnosing glaucoma.

Methods
Healthy eyes (n=102) and glaucomatous eyes (n=101) underwent reliable Oculus Spark perimetry, and imaging with the HRT, GDx, and Cirrus OCT. In addition, ONH color images were acquired with a non-mydriatic fundus camera. The Laguna ONhE program then was used to calculate the Hb amount in each of 24 sectors of the ONH. Sensitivities at 95% fixed specificity, diagnostic agreement, and linear correlations between parameters with the best diagnostic ability were calculated.

Results
The glaucoma discriminant function (GDF) of the Laguna program, evaluating Hb in the vertical intermediate sectors and center/periphery Hb amount slope, yielded an 89.1% sensitivity and 95.1% specificity, which was superior or similar to the other tests. The best GDF diagnostic agreement was for the OCT-vertical cup-to-disc (C/D) ratio (kappa = 0.772) and the final phase Spark pattern SD (kappa = 0.672).

Hb levels correlated strongly with the Spark mean sensitivity (first phase 0.70, final phase 0.71). Hb also correlated well with the Reinhard OW Burk discriminant function of the HRT (0.56), nerve fiber indicator of GDx (0.64), and vertical C/D ratio of OCT (0.71).

Read publication

Optic Nerve Head hemoglobin levels in childhood glaucoma patients

6th World Glaucoma Congress 2015 - Abstract Book
DOI:
Carmen Mendez-Hernandez, Lucia Perucho, Julian Garcia Feijoo, Manuel Gonzalez de la Rosa

Purpose
The computer program Laguna ONhE determines optic nerve head hemoglobin (ONH Hb) on retinal photographs based on detecting colour differences. The software provides two diagnostic indices for glaucoma: estimated vertical cup-disc-ratio (C/D) and glaucoma discriminant function GDF). This study examines the amount of ONH Hb in patients with chilhood glaucoma using this new noninvasive technique.

Methods
In this prospective, observational case series study, measurements were made on retinal photographs (Canon CR-Dgi non mydriatic fundus camera) using the Laguna ONhE program in 108 eyes of 63 healthy subjects and 88 eyes of 56 patients with childhood glaucoma. The variables recorded were: C/D, GDF, and ONH Hb across the whole disc, and across the vertical disc diameter (sectors 8 and 20). ONH Hb differences between groups were determined by independent t Student test. U Mann Whitney test was used in non parametric parameters. Pearson’s correlation and lineal regression model were assessed in both childhood glaucoma and control study group.

Results
The median age in childhood glaucoma was 14 years old (P25-P75 10;25) and 9 years old (P25-P75 7;13) in healthy subjects (p 0.000).ONH Hb across vertical disc diameter was higher in controls (64.62 ± 7.52%) than in glaucomatous eyes (59.96 ± 13.07%), p0.002. C/D was higher in glaucomatous eyes (0.61 ± 0.17) than in control eyes (0.52 ± 0.98), p0.000. GDF was lower in glaucoma (-4 P25-P75 -30;20) than in the control group (6 P25-P75 -2;19), p0.001. There were not significant differences in ONH Hb across the whole disc between childhood glaucoma eyes (57.75± 11.24%) and healthy eyes (58.14 ± 7.16%) p 0.770. C/D on glaucoma patients was correlated with ONH Hb across the whole disc (- 0.745, p 0.000), ONH Hb across the vertical disc diameter (- 0.885, p 0.000) and GDF index (- 0.981, p 0.000). Multiple linear regression analysis revealed an effect of age (slope –0.153%/year (95%CI -0.61; -0.02, p = 0.023) on ONH Hb.

Conclusion
Our findings indicate the capacity of this device in childhood glaucoma diagnosis, however ONH Hb across the whole disc may have normal values. Our results will help to make future adjustments to the software of this new program.

Read publication

Estimating the Amount of Hemoglobin in the Neuroretinal Rim Using Color Images and OCT

Current Eye Research
Erica Medina-Mesa, Marta Gonzalez-Hernandez, Jose Sigut, Francisco Fumero-Batista, Cristina Pena-Betancor, Silvia Alayon & Manuel Gonzalez dela Rosa

The amount of Hb in the ONH has an important relationship with glaucomatous eyes

This study assessed the amount of Hb (hemoglobin) in healthy and glaucomatous eyes. The amount of Hb in the ONH seems to have an important relationship with glaucomatous visual field sensitivity. 

Purpose
To calculate the amount of hemoglobin (Hb) in the optic nerve head (ONH), using superimposed color fundus images with disc, rim and cup boundaries obtained by OCT-Cirrus.

Methods
We examined 100 healthy and 121 glaucomatous eyes using Oculus–Spark perimetry, Cirrus-OCT and Visucam (Zeiss) ONH color images. The Laguna ONhE program was then used to calculate the amount of Hb in the cup and six sectors of the rim. Receiver operating characteristic (ROC) analysis was performed and correlations between parameters were calculated.

Results
In suspected and confirmed glaucoma, Hb was significantly lower than controls in all rim sectors, especially the inferior and superonasal (p < 0.0001). Mean deviation (MD) of visual field regions showed greater correlation with the amount of Hb in the superior and inferior sectors of the rim than with rim area (p = 0.02) or nerve fiber layer thickness (p < 0.0001). On ROC analysis, the best diagnostic indicators were OCT rim area, vertical cup/disc ratio (C/D) and Glaucoma Discriminant Function (GDF) of Laguna ONhE, without significant differences.

Conclusions
The amount of Hb in the ONH seems to have an important relationship with glaucomatous visual field sensitivity. The remaining rim has insufficient perfusion in many cases of glaucoma.

Read publication

Glaucoma diagnostic capacity of optic nerve head haemoglobin measures compared with spectraldomain OCT and HRT III confocal tomography

Acta Ophthalmologica
Carmen Mendez-Hernandez, Ignacio Rodriguez-Una, Manuel Gonzalez-de-la Rosa, Paula Arribas-Pardo and Julian Garcia-Feijoo

Similar diagnostic power for RetinaLyze Glaucoma, an OCT and a confocal tomography

This study compares the diagnostic capacity of the RetinaLyze Glaucoma algorithm with that of spectral domain optical coherence tomography (OCT) and confocal tomography (HRT III) and found similar diagnostic power.

Purpose
The computer program laguna onhe determines optic nerve head haemoglobin (ONH Hb) on retinal photographs based on detecting colour differences. This study compares the diagnostic capacity of Laguna ONhE with that of spectral domain optical coherence tomography (OCT) and confocal tomography (HRT III).

Methods
In a prospective, observational, cross-sectional study, glaucomatous (n = 66) and healthy (n = 52) eyes were examined by Spectralis OCT, HRT III and Laguna ONhE. The following Laguna ONhE variables were determined: ONH Hb across the vertical disc diameter (8&20 Hb), estimated cup-disc ratio (C/D) and the glaucoma discriminant function (GDF), which combines the slope of Hb amount with the mean in 8&20 Hb. The three diagnostic methods were compared by calculating areas under ROC curves (AUCs). Correlations between variables were assessed through Spearman's rho coefficient.

Results
Areas under ROC curves (AUCs) were 0.785 (95% CI: 0.700-0.863) for GDF, 0.807 (95% CI: 0.730-0.883) for OCT retinal nerve fibre layer thickness (OCT-RNFL) and 0.714 (95% CI: 0.618-0.810) and 0.721 (95% CI: 0.628-0.815) for the HRT III variable GPS (glaucoma probability score) and vertical C/D ratio, respectively. Glaucoma discriminant function (GDF) was correlated with OCT-RNFL (0.587, p 0.001; 0.507, p 0.045; and -0.119, p 0.713 for mild, moderate and advanced glaucoma, respectively), mostly so with inferior OCT-RNFL (0.622; p < 0.001). Glaucoma discriminant function (GDF)-HRT III correlations were lower (rim area 0.471, p < 0.0001; rim/disc area 0.426, p < 0.0001; vertical C/D -0.413, p < 0.0001; GPS -0.408, p < 0.0001; rim volume 0.341, p < 0.0001).

Conclusion
Similar diagnostic power was observed for Laguna ONhE, Spectralis OCT and HRT III.

Read publication

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

Acta Ophthalmologica
A. Hansen, N. Hartvig, M. Jensen, K. Borch-Johnsen, H. Lund-Andersen, M. Larsen

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.

Read publication

Automated Detection of Diabetic Retinopathy: Results of a Screening Study

Diabetes Technology & Therapeutics
Bouhaimed Manal, Gibbins Robbie, and Owens David

Purpose
This study evaluated the operating characteristics of a reading software (Retinalyze® System, Retinalyze A/S, Hørsholm, Denmark) for automated prescreening of digital fundus images for diabetic retinopathy.

Methods
Digital fundus images of patients with diabetes were retrospectively selected from the Bro Taf diabetic retinopathy screening program in Wales, UK in the period of 2002–2004, which has been superseded by the Diabetic Retinopathy Screening Service for Wales. A gold standard reference was defined by classifying each patient as having or not having diabetic retinopathy based on overall visual grading of the digitized images using the Bro Taf reading protocol. Automated grading was applied using automated red or bright lesion detection at varying detection sensitivities and adjusting for image quality. Operating characteristics included sensitivity, specificity, positive predictive values, and negative predictive values (PPV and NPV, respectively).

Results
Automated analysis of four hundred fundus photographs of 192 eyes from 96 patients with diabetes was performed. The automated red lesion detection had a sensitivity of 82%, specificity of 75%, PPV of 41%, and NPV of 95%. Combined automated red and bright lesion detection yielded a sensitivity of 88%, specificity of 52%, PPV of 28%, and NPV of 95%. Performance of the combined red and bright lesion detection at elevated thresholds in images of good quality demonstrated a sensitivity of 93%, specificity of 78%, PPV of 46%, and NPV of 98%.

Conclusions
Prescreening for diabetic retinopathy by automated detection of single fundus lesions seem to be achieved with minimal false negativity and can help to decrease the burden of manual diabetic retinopathy screening.

Read publication

Automated Detection of Diabetic Retinopathy in a Fundus Photographic Screening Population

Investigative Ophthalmology & Visual Science
Nicolai Larsen, Jannik Godt, Michael Grunkin, Henrik Lund-Andersen and Michael Larsen

Evaluation of RetinaLyze

RetinaLyze allied with the University of Copenhagen, the Steno Diabetes Centre and Herlev Hospital to further evaluate the system.

Purpose
To evaluate the performance of an automated fundus photographic image-analysis algorithm in high-sensitivity and/or high-specificity segregation of patients with diabetes with untreated diabetic retinopathy from those without retinopathy.

Methods
This was a retrospective cross-sectional study of 260 consecutive nonphotocoagulated eyes in 137 diabetic patients attending routine photographic retinopathy screening. Mydriatic 60° fundus photography on 35-mm color transparency film was used, with a single fovea-centered field. Routine grading was based on visual examination of slide-mounted transparencies. Reference grading was performed with specific emphasis on achieving high sensitivity. Computer-assisted automated red lesion detection was performed on digitized transparencies.

Results
When applied in a screening population comprising patients with diabetes with untreated diabetic retinopathy in any eye and patients with diabetes without retinopathy, the automated lesion detection correctly identified 90.1% of patients with retinopathy and 81.3% of patients without retinopathy. A per-eye analysis for methodological purposes demonstrated that the automated lesion detection could be adapted to simulate various visual evaluation strategies. When adapted at high sensitivity, the automated system demonstrated sensitivity at 93.1% and specificity at 71.6%. When adapted at high specificity the automated system demonstrated sensitivity at 76.4% and specificity at 96.6%, closely matching routine visual grading at sensitivity 76.4% and specificity 98.3%.

Conclusions
Automated detection of untreated diabetic retinopathy in fundus photographs from a screening population of patients with diabetes can be made with adjustable priority settings, emphasizing high-sensitivity identification of diabetic retinopathy or high-specificity identification of absence of retinopathy, covering opposing extremes of visual evaluation strategies demonstrated by human observers.

Read publication

The efficacy of automated ‘‘disease/no disease’’ grading for diabetic retinopathy in a systematic screening programme

British Journal of Ophthalmology
S. Philip, A. D. Fleming, K. A. Goatman, S. Fonseca, P. Mcnamee, G. S. Scotland, G. J. Prescott, P. F. Sharp, J. A. Olson

Safe and secure

The British Journal of Ophthalmology published a report which described the work of several renowned scientist who investigated if digital automated detection of diabetic retinopathy could be used for a systematic and safe screening programme.

Purpose
To assess the efficacy of automated “disease/no disease” grading for diabetic retinopathy within a systematic screening programme.

Methods
Anonymised images were obtained from consecutive patients attending a regional primary care based diabetic retinopathy screening programme. A training set of 1067 images was used to develop automated grading algorithms. The final software was tested using a separate set of 14 406 images from 6722 patients. The sensitivity and specificity of manual and automated systems operating as “disease/no disease” graders (detecting poor quality images and any diabetic retinopathy) were determined relative to a clinical reference standard.

Results
The reference standard classified 8.2% of the patients as having ungradeable images (technical failures) and 62.5% as having no retinopathy. Detection of technical failures or any retinopathy was achieved by manual grading with 86.5% sensitivity (95% confidence interval 85.1 to 87.8) and 95.3% specificity (94.6 to 95.9) and by automated grading with 90.5% sensitivity (89.3 to 91.6) and 67.4% specificity (66.0 to 68.8). Manual and automated grading detected 99.1% and 97.9%, respectively, of patients with referable or observable retinopathy/maculopathy. Manual and automated grading detected 95.7% and 99.8%, respectively, of technical failures.

Conclusion
Automated “disease/no disease” grading of diabetic retinopathy could safely reduce the burden of grading in diabetic retinopathy screening programmes.

Read publication

Automated Detection of Fundus Photographic Red Lesions in Diabetic Retinopathy

Investigative Ophthalmology & Visual Science, February 2003, Vol. 44, No. 2 Copyright © Association for Research in Vision and Ophthalmology
Michael Larsen, Jannik Godt, Nicolai Larsen, Henrik Lund-Andersen, Anne Katrin Sjølie, Elisabet Agardh, Helle Kalm, Michael Grunkin, and David R. Owens.

Matches manual detection

Retinalyze allied with specialists from the University of Copenhagen, Herlev Hospital, Odense University Hospital, Malmö University Hospital, Sahlgrenska University Hospital and the University of Wales College of Medicine to find out if the performance of Retinalyze could match the performance of experienced ophthalmologists.

Purpose
To compare a fundus image-analysis algorithm for automated detection of hemorrhages and microaneurysms with visual detection of retinopathy in patients with diabetes.

Methods
Four hundred fundus photographs (35-mm color transparencies) were obtained in 200 eyes of 100 patients with diabetes who were randomly selected from the Welsh Community Diabetic Retinopathy Study. A gold standard reference was defined by classifying each patient as having or not having diabetic retinopathy based on overall visual grading of the digitized transparencies. A single-lesion visual grading was made independently, comprising meticulous outlining of all single lesions in all photographs and used to develop the automated red lesion detection system. A comparison of visual and automated single-lesion detection in replicating the overall visual grading was then performed.

Results
Automated red lesion detection demonstrated a specificity of 71.4% and a resulting sensitivity of 96.7% in detecting diabetic retinopathy when applied at a tentative threshold setting for use in diabetic retinopathy screening. The accuracy of 79% could be raised to 85% by adjustment of a single user-supplied parameter determining the balance between the screening priorities, for which a considerable range of options was demonstrated by the receiver-operating characteristic (area under the curve 90.3%). The agreement of automated lesion detection with overall visual grading (0.659) was comparable to the mean agreement of six ophthalmologists (0.648).

Conclusions
Detection of diabetic retinopathy by automated detection of single fundus lesions can be achieved with a performance comparable to that of experienced ophthalmologists. The results warrant further investigation of automated fundus image analysis as a tool for diabetic retinopathy screening.

Read publication

A health economic analysis

The Danish Committee for Health Education
DOI:
Michael Nyhus Andreasen, Jacob Kjellberg. The Danish Committee for Health Education

Cut costs and time

DSI (now KORA), the Danish Institute of Health, and Signifikans, a CRO in statistical analysis, concluded that RetinaLyze could severely decrease the time and money spent on manual grading in societies.

Purpose
To implement an operational economic evaluation of the Retinalyze software system used for the automated detection of the diabetic eye disease, retinopathy, by means of fundus photography.

Methods
Review of the present relevant literature. Comparison and analysis of RetinaLyze-related studies.

Results
The marginal cost per screening is between DKK 13.10 and DKK 27.16 with the automated analysis instead of the visual. At e.g. annual screenings at a typical eye doctor’s practice, the costs total between DKK 60,800 and DKK 87,300, varying with the number of patients screened positive. This presupposes internalization of savings achieved through task-shifting (time release for doctors).
The total costs must be compared to an expected activity progress in the number of eye screenings caused by the task-shifting and the related incomes in the form of e.g. health insurance services. In addition to this, there are non-quantified gains in the form qualitative implications of the screenings, improved opportunities for time-series studies accompanied by image database development as well as improved opportunities for telescreening.
With conservative model assumptions, about 11.1 full-time eye doctor positions are used on examinations per year in connection with diabetic retinopathy, with the present number of diabetics and with the use of ophthalmoscopy and/or visual image analysis.
With the use of the Retinalyze software system for automated image analysis, the task-shifting will influence the number of eye doctor positions used for examinations in connection with diabetic retinopathy on the decline and causes a saving potential in the form of released time corresponding to 8.4 full-time eye doctor position per year, based on conservative conditions.
Sensitivity analysis do not change the conclusion of the analysis significantly, as the saving potential obtained as time release is still estimated to be important in a situation with eye doctor shortage, relatively long waiting lists and an expected increase in the demand for screenings.

Conclusions
With the review of the present literature, it is concluded that the detection of diabetic retinopathy by means of automated image analysis can be achieved with a system performance directly comparable 19 to the one for experienced ophthalmologists with visual image analysis. The results prove further examination of automated fundus image analysis as a screening tool. Automated image analysis instead of visual image analysis causes a task-shifting, with a shift of tasks from ophthalmologists to nurses / specially trained paramedical personnel.

Read publication

Assessment of Automated Screening for Treatment-Requiring Diabetic Retinopathy

Current Eye Research, Informa Healthcare
Michael Larsen, Theis Gondolf, Jannik Godt, Maja Skytte Jensen, Niels Væver Hartvig, Henrik Lund-Andersen, and Nicolai Larsen. Department of Ophthalmology, Glostrup Hospital, University of Copenhagen, Glostrup, Denmark

Sensitivity: Up to 100%

The Department of Ophthalmology, Rigshospitalet – Glostrup and University of Copenhagen, Denmark conducted a study of the Retinalyze software and concluded that the software could be used as a first-step screening tool and their results showed that the software had a sensitivity of 100%, exact CI95 96.6–100.0%. Their study concluded that the combination of an automated screening algorithm and an automated picture quality assessment was very effective.

Purpose
To evaluate fundus photographic image analysis combining automated detection of red lesions, bright lesions, and image quality as a means of identifying treatment-requiring diabetic retinopathy in a screening population of diabetic patients. 

Methods
This was a retrospective cross-sectional study of 106 patients from a diabetic retinopathy screening clinic referred for photocoagulation treatment in the period from January 1996 to May 2002 on the basis of mydriatic 60-degree 35-mm color transparency fundus photography. One fovea-centered fundus photograph and one centered nasal of the optic disk from each of a subject’s two eyes was selected for digitization and analyzed using a previously tested computerized red-lesion detection algorithm in combination with a new algorithm for detection of bright lesions and image quality. The algorithm was calibrated on an independent set of fundus photographs. 

Results
Automated red-lesion detection identified 104 of 106 patients requiring photocoagulation treatment, whereas bright-lesion detection identified only 91 of the 106 patients. Two patients who were not identified by either lesion detection algorithm were automatically detected as having poor image quality in one or both eyes. In the study sample, the risk of missing treatment-requiring retinopathy patients from being detected was 0.0% (estimated CI95 0.0–3.4%). 

Conclusions
The combination of automated detection of red lesions and poor image quality identified all treatment-requiring diabetic retinopathy patients in the study sample. No additional information was contributed by the automated bright-lesion detection.

Read publication