Carlos Manta Oliveira

Microaneurysm Formation Rate As A Predictive Marker For Progression To Clinically Significant Macular Edema In Nonproliferative Diabetic Retinopathy

Purpose: To evaluate the predictive value of microaneurysm (MA) formation rate concerning the development of clinically significant macular edema (CSME) in patients with mild-to-moderate nonproliferative diabetic retinopathy as evaluated by an automated analysis of central field fundus 30° photographs. Methods: Two hundred and eighty-seven eyes were included in the study. Photographs obtained at Day 0, at 6, and 12 months were analyzed using the RetmarkerDR software (Critical Health SA) in a masked manner, and the MA formation rate was documented. A threshold of a calculated MA formation rate of 2 or more was chosen to consider a patient “positive.” The ability to predict CSME development was then calculated for a period of up to 5 years. HbA1c values, blood pressure, or duration of diabetes were also evaluated. Results: The study population consisted of 89 male and 59 female patients with a mean age of 57.6 years, a mean HbA1c of 7.8, and a mean duration of diabetes of 12.3 years. Forty-seven of 287 eyes (16.4%) developed CSME during follow-up. An increased MA formation rate of >2 MA was clearly associated with development of CSME. Using the automated analysis and a threshold of 2 or more new MA, the authors were able to identify 70.2% of the eyes that developed CSME during follow-up (“true positive”) and using a threshold of up to 2 new MA, 71.7% of the patients that did not develop CSME (“true negative”). No significant differences concerning baseline and 1-year HbA1c levels within patient eyes that developed CSME compared with patient eyes below or over the calculated threshold of 2 MA (P = 0.554 and P = 0.890, respectively) were seen. The positive and negative predictive value was calculated to be 33% versus 92.5%, sensitivity was 70%, and specificity was 72%. Conclusion: Using the RetmarkerDR software, the authors were able to identify patients with higher risk to develop CSME during follow-up using a threshold of 2 or more MA formation rate. Together with the high negative predictive value, the automated analysis may help to determine the individual risk of a patient to develop sight-threatening complications related to diabetic retinopathy and schedule individual screening intervals.

Improved Automated Screening of Diabetic Retinopathy

Aim: To assess a two-step automated system (RetmarkerSR) that analyzes retinal photographs to detect diabetic retinopathy for the purpose of reducing the burden of manual grading. Methods: Anonymous images from 5,386 patients screened in 2007 were obtained from a nonmydriatic diabetic retinopathy screening program in Portugal and graded by an experienced ophthalmologist. RetmarkerSR earmarked microaneurysms, generating two outputs: ‘disease’ or ‘no disease’. A second-step analysis, based on coregistration, combining two visits, was subsequently performed in 289 patients who underwent repeated examinations in 2008. The study was extended by analyzing all referrals considered urgent by the ophthalmologist from 2001 to 2007. Results were compared with those obtained by manual grading. Results: The RetmarkerSR classified in a first-step analysis 2,560 patients (47.5%) as having ‘no disease’ and 2,826 patients (52.5%) as having ‘disease’, thus requiring manual grading. RetmarkerSR detected all eyes considered urgent referrals. The two-step analysis further reduced the number of false-positive results by 26.3%, indicating an overall sensitivity of 95.8% and a specificity of 63.2%. Conclusion: Automated grading of diabetic retinopathy may safely reduce the burden of grading patients in diabetic retinopathy screening programs. The novel two-step automated analysis system offers improved sensitivity and specificity over published automated analysis systems.

Automated lesion detectors in retinal fundus images

Diabetic retinopathy (DR) is a sight-threatening condition occurring in persons with diabetes, which causes progressive damage to the retina. The early detection and diagnosis of DR is vital for saving the vision of diabetic persons. The early signs of DR which appear on the surface of the retina are the dark lesions such as microaneurysms (MAs) and hemorrhages (HEMs), and bright lesions (BLs) such as exudates. In this paper, we propose a novel automated system for the detection and diagnosis of these retinal lesions by processing retinal fundus images. We devise appropriate binary classifiers for these three different types of lesions. Some novel contextual/numerical features are derived, for each lesion type, depending on its inherent properties. This is performed by analysing several wavelet bands (resulting from the isotropic undecimated wavelet transform decomposition of the retinal image green channel) and by using an appropriate combination of Hessian multiscale analysis, variational segmentation and cartoon+texture decomposition. The proposed methodology has been validated on several medical datasets, with a total of 45,770 images, using standard performance measures such as sensitivity and specificity. The individual performance, per frame, of the MA detector is 93% sensitivity and 89% specificity, of the HEM detector is 86% sensitivity and 90% specificity, and of the BL detector is 90% sensitivity and 97% specificity. Regarding the collective performance of these binary detectors, as an automated screening system for DR (meaning that a patient is considered to have DR if it is a positive patient for at least one of the detectors) it achieves an average 95-100% of sensitivity and 70% of specificity at a per patient basis. Furthermore, evaluation conducted on publicly available datasets, for comparison with other existing techniques, shows the promising potential of the proposed detectors.

Ocular Risk Factors for Exudative AMD: A Novel Semiautomated Grading System

Purpose. To evaluate the contribution of the ocular risk factors in the conversion of the fellow eye of patients with unilateral exudative AMD, using a novel semiautomated grading system. Materials and Methods. Single-center, retrospective study including 89 consecutive patients with unilateral exudative AMD and ≥3 years of followup. Baseline color fundus photographs were graded using an innovative grading software, RetmarkerAMD (Critical Health SA). Results. The follow-up period was 60.9 ± 31.3 months. The occurrence of CNV was confirmed in 42 eyes (47.2%). The cumulative incidence of CNV was 23.6% at 2 years, 33.7% at 3 years, 39.3% at 5 years, and 47.2% at 10 years, with a mean annual incidence of 12.0% (95% CI = 0.088–0.162). The absolute number of drusen in the central 1000 and 3000 μm (P < 0.05) and the absolute number of drusen ≥125 µm in the central 3000 and 6000 µm (P < 0.05) proved to be significant risk factors for CNV. Conclusion. The use of quantitative variables in the determination of the OR of developing CNV allowed the establishment of significant risk factors for neovascularization. The long follow-up period and the innovative methodology reinforce the value of our results. This trial is registered with ClinicalTrials.gov NCT00801541.

Microaneurysm Turnover in Diabetic Retinopathy Assessed by Automated RetmarkerDR Image Analysis - Potential Role as Biomarker of Response to Ranibizumab Treatment

Purpose: To evaluate the influence of a ranibizumab treatment on microaneurysm (MA) turnover in diabetic retinopathy. Methods: Sixty-nine eyes were included in this retrospective study. We compared a group of 33 eyes with ranibizumab treatment for diabetic macular edema to 36 eyes with nonproliferative diabetic retinopathy only. Nonmydriatic ultra-widefield scanning laser ophthalmoscopy (Optomap) images were obtained at a mean 4.76 ± 1.69 days prior to the first ranibizumab injection (baseline) and again 35.94 ± 2.44 days after the third consecutive injection in a 4-week interval. In untreated controls, images were obtained at baseline and 97.81 ± 3.16 days thereafter. Images were analyzed using the RetmarkerDR software (Critical Health SA, Coimbra, Portugal), and the turnover of MAs was documented and analyzed. Thereafter, MA turnover was correlated with central retinal thickness (CRT) as assessed by OCT. Results: At baseline, patients in the treatment group had 5.64 ± 0.75 MAs. One month after 3 ranibizumab injections, measured MAs decreased to 4.03 ± 0.66. In the untreated control group, the initial number of 3.36 ± 0.6 MAs remained almost unchanged over 3-4 months (2.89 ± 0.57 MAs). Dynamic analysis showed that after ranibizumab treatment 3.06 ± 0.5 new MAs appeared, while 5.09 ± 0.79 disappeared. In the control group, 2.11 ± 0.4 new MAs appeared and 2.61 ± 0.48 disappeared. MA turnover was significantly higher with ranibizumab compared to the control group (8.15 ± 1.14 vs. 4.72 ± 0.81, p < 0.001). Consistently, CRT decreased from 444 to 330 µm in the ranibizumab group, while there was no change in the control group (291 vs. 288 µm). Conclusion: The treatment of macular edema using ranibizumab does not only reduce macular thickness, but also has an impact on the turnover of MAs in diabetic retinopathy. RetmarkerDR analysis showed that more pre-existent MAs disappeared than new MAs developed, and the absolute number of MAs also decreased.

Screening for Diabetic Retinopathy in the Central Region of Portugal. Added Value of Automated 'Disease/No Disease' Grading.

Purpose: To describe the procedures of a nonmydriatic diabetic retinopathy (DR) screening program in the Central Region of Portugal and the added value of the introduction of an automated disease/no disease analysis. Methods: The images from the DR screening program are analyzed in a central reading center using first an automated disease/no disease analysis followed by human grading of the disease cases. The grading scale used is as follows: R0 - no retinopathy, RL - nonproliferative DR, M - maculopathy, RP - proliferative DR and NC - not classifiable. Results: Since the introduction of automated analysis in July 2011, a total of 89,626 eyes (45,148 patients) were screened with the following distribution: R0 - 71.5%, RL - 22.7%, M - 2.2%, RP - 0.1% and NC - 3.5%. The implemented automated system showed the potential for human grading burden reduction of 48.42%. Conclusions: Screening for DR using automated analysis allied to a simplified grading scale identifies DR vision-threatening complications well while decreasing human burden.

Computer-Aided Detection of Diabetic Retinopathy Progression

It is considered crucial for diabetic retinopathy (DR) management to identify disease progression in clinical practice. Automated computer-aided analysis of fundus digital photographs giving microaneurysm formation and disappearance rates together with OCT measurements of extracellular space and retinal thickness, both based on non-invasive procedures allow close follow-up of the main alterations occurring in the diabetic retina: microaneurysm turnover, capillary closure and alteration of blood-retinal barrier. Determination of the activity of the retinal disease and individual risk profiles using non-invasive procedures is expected to contribute to personalized management of diabetic retinopathy and prevent its vision-threatening complications, macular oedema and proliferative retinopathy. Finally, automated computer-aided analysis of fundus digital photographs, namely, the Retmarker, offers a promising contribution to reduce the burden of manual grading in DR screening programmes.

Pattern Classes in Retinal Fundus Images Based on Function Norms

Retinal fundus images are widely used for screening, diagnosis and prognosis purposes in ophthalmology. Additionally these can also be used in retinal identification/recognition systems, for identification/authentication of an identity. In this paper the aim is to explain how norms in function spaces can be used to set up, automatically, classes of different retinal fundus images. These classifications rely on crucial and unique retinal features, such as the vascular network, whose location and measurement are appropriately quantified by weighted norms in function spaces. These quantifications can be understood as retinal pattern assessments and used for improving the efficiency and speed of retinal identification/recognition frameworks. The proposed methods are evaluated in a large dataset of retinal fundus images, and, besides being very fast, they achieve a reduction of the search in the dataset (for identification/recognition purposes), by 70% on average.

Automated retina identification based on multiscale elastic registration

In this work we propose a novel method for identifying individuals based on retinal fundus image matching. The method is based on the image registration of retina blood vessels, since it is known that the retina vasculature of an individual is a signature, i.e., a distinctive pattern of the individual. The proposed image registration consists of a multiscale affine registration followed by a multiscale elastic registration. The major advantage of this particular two-step image registration procedure is that it is able to account for both rigid and non-rigid deformations either inherent to the retina tissues or as a result of the imaging process itself. Afterwards a decision identification measure, relying on a suitable normalized function, is defined to decide whether or not the pair of images belongs to the same individual. The method is tested on a data set of 21721 real pairs generated from a total of 946 retinal fundus images of 339 different individuals, consisting of patients followed in the context of different retinal diseases and also healthy patients. The evaluation of its performance reveals that it achieves a very low false rejection rate (FRR) at zero FAR (the false acceptance rate), equal to 0.084, as well as a low equal error rate (EER), equal to 0.053. Moreover, the tests performed by using only the multiscale affine registration, and discarding the multiscale elastic registration, clearly show the advantage of the proposed approach. The outcome of this study also indicates that the proposed method is reliable and competitive with other existing retinal identification methods, and forecasts its future appropriateness and applicability in real-life applications.

Automatic detection of laser marks in retinal digital fundus images

Diabetic retinopathy (DR) is the most frequent complication of diabetes mellitus that affects vision to the point of causing blindness. In advanced stages its progress can be delayed with laser photocoagulation which leaves behind marks on the retina. Modern screening programs rely on automatic diagnostic algorithms to detect signs of DR in patients. These systems performance may be impaired when patient retina presents marks from previous laser photocoagulation treatments. Since these patients are already being treated, it is desirable to detect and remove them from the screening program. An algorithm that automatically detects the presence of laser marks in retinal images using tree-based classifiers is proposed and the results on its performance are obtained and described. Two new public accessible datasets containing retinal images with laser marks are provided in this paper.