Age-related macular degeneration (AMD) is a common threat to vision. While classification of disease stages is critical to understanding disease risk and progression, several systems based on color fundus photographs are known. Most of these require in-depth and time-consuming analysis of fundus images. Herein, we present an automated computer-based classification algorithm. Algorithm development for AMD classification based on a large collection of color fundus images. Validation is performed on a cross-sectional, population-based study. We included 120 656 manually graded color fundus images from 3654 Age-Related Eye Disease Study (AREDS) participants. AREDS participants were >55 years of age, and non-AMD sight-threatening diseases were excluded at recruitment. In addition, performance of our algorithm was evaluated in 5555 fundus images from the population-based Kooperative Gesundheitsforschung in der Region Augsburg (KORA; Cooperative Health Research in the Region of Augsburg) study. We defined 13 classes (9 AREDS steps, 3 late AMD stages, and 1 for ungradable images) and trained several convolution deep learning architectures. An ensemble of network architectures improved prediction accuracy. An independent dataset was used to evaluate the performance of our algorithm in a population-based study. κ Statistics and accuracy to evaluate the concordance between predicted and expert human grader classification. A network ensemble of 6 different neural net architectures predicted the 13 classes in the AREDS test set with a quadratic weighted κ of 92% (95% confidence interval, 89%-92%) and an overall accuracy of 63.3%. In the independent KORA dataset, images wrongly classified as AMD were mainly the result of a macular reflex observed in young individuals. By restricting the KORA analysis to individuals >55 years of age and prior exclusion of other retinopathies, the weighted and unweighted κ increased to 50% and 63%, respectively. Importantly, the algorithm detected 84.2% of all fundus images with definite signs of early or late AMD. Overall, 94.3% of healthy fundus images were classified correctly. Our deep learning algoritm revealed a weighted κ outperforming human graders in the AREDS study and is suitable to classify AMD fundus images in other datasets using individuals >55 years of age.

Local, abdominal fat depots may be related to alterations in cardiac function and morphology due to a metabolic linkage. Thus, we aimed to determine their association with subtle cardiac changes and the potential interaction with hyperglycemic metabolic states. Subjects from the general population and without history of cardiovascular disease were drawn from the Cooperative Health Research in the Region of Augsburg FF4 cohort and underwent 3 T cardiac and body MRI. Measures of abdominal adiposity such as hepatic proton-density fat fraction [PDFFhepatic], subcutaneous (SAT) and visceral abdominal fat (VAT) as well as established cardiac left-ventricular (LV) measures including LV remodeling index (LVCI) were derived. Associations were determined using linear regression analysis based on standard deviation normalized predictors. Among a total of 374 subjects (56.2 ± 9.1 years, 58% males), 49 subjects had diabetes, 99 subjects had prediabetes and 226 represented normal controls. Only subtle cardiac alterations were observed (e.g. LVCI: 1.13 ± 0.30). While SAT was not associated, increasing VAT and increasing PDFFhepatic were independently associated with increasing LVCI (β = 0.11 and 0.06, respectively), decreasing LV end-diastolic volume (β = − 6.70 and 3.23, respectively), and decreasing LV stroke volume (β = − 3.91 and − 2.20, respectively). Hyperglycemic state did not modify the associations between VAT or PDFF and LV measures (interaction term: all p ≥ 0.29). In a healthy population, VAT but also PDFFhepatic were associated with subclinical measures of LV remodeling without evidence for a modifying effect of hyperglycemic state.

Left ventricular (LV) hypertrophy and changes in LV geometry are associated with increased cardiovascular mortality. Subjects with type 2 diabetes have an increased risk of such alterations in cardiac morphology. We sought to assess the association of glycemic status and LV wall thickness measured by cardiac magnetic resonance (CMR), and potential interactions of hypertension and diabetes. CMR was performed on 359 participants from a cross-sectional study nested in a population-based cohort (KORA FF4) free of overt cardiovascular disease. Participants were classified according to their glycemic status as either control (normal glucose metabolism), prediabetes or type 2 diabetes. Segmentation of the left ventricle was defined according to the American Heart Association (AHA) 16-segment model. Measurements of wall thickness were obtained at end-diastole and analyzed by linear regression models adjusted for traditional cardiovascular risk factors. LV wall thickness gradually increased from normoglycemic controls to subjects with prediabetes and subjects with diabetes (8.8 ± 1.4 vs 9.9 ± 1.4 vs 10.5 ± 1.6 mm, respectively). The association was independent of hypertension and traditional cardiovascular risk factors (β-coefficient: 0.44 mm for prediabetes and 0.70 mm for diabetes, p-values compared to controls: p = 0.007 and p = 0.004, respectively). Whereas the association of glycemic status was strongest for the mid-cavity segments, the association of hypertension was strongest for the basal segments. Abnormal glucose metabolism, including pre-diabetes, is associated with increased LV wall thickness independent of hypertension.

Aims/hypothesis: The prevalence of type 2 diabetes is increasing worldwide, and previous studies have suggested that it is higher in individuals who are seropositive for herpesviruses. This study examines the prospective association of herpesviruses with (pre)diabetes to evaluate their potential role in diabetes aetiology. Methods: Two follow-up examinations of the German population-based KORA cohort (F4 and FF4) were used to identify participants with normal glucose tolerance at baseline, thus being at risk for (pre)diabetes (n = 1257). All participants had repeated OGTTs and antibody measurements for herpes simplex virus (HSV) 1 and 2, varicella-zoster virus, Epstein–Barr virus, cytomegalovirus (CMV) and human herpesvirus 6 and 7. Regression models were used to evaluate the association between serostatus with (pre)diabetes incidence after a 7 year follow-up and HbA_1c. Results: HSV2 and CMV were associated with (pre)diabetes incidence after adjustment for sex, age, BMI, education, smoking, physical activity, parental diabetes, hypertension, lipid levels, insulin resistance and fasting glucose. Seropositivity of both viruses was also cross-sectionally associated with higher HbA_1c at baseline, with the association of HSV2 being independent of confounders, including the prevalence of (pre)diabetes itself. While seropositivity for multiple herpesviruses was associated with a higher incidence of (pre)diabetes, this association was not independent of confounders. Conclusions/interpretation: The associations of HSV2 and CMV serostatus with (pre)diabetes incidence indicate that these herpesviruses may contribute to the development of impaired glucose metabolism. Our results highlight the link between viral infection and (pre)diabetes, and the need for more research evaluating viral prevention strategies. Graphical