Diabetic retinopathy (DR) is a common retinal disease that leads to blindness. For diagnosis purposes, DR image grading aims to provide automatic DR grade classification, which is not addressed in conventional research methods of binary DR image classification. Small objects in the eye images, like lesions and microaneurysms, are essential to DR grading in medical imaging, but they could easily be influenced by other objects. To address these challenges, we propose a new deep learning architecture, called BiRA-Net, which combines the attention model for feature extraction and bilinear model for fine-grained classification. Furthermore, in considering the distance between different grades of different DR categories, we propose a new loss function, called grading loss, which leads to improved training convergence of the proposed approach. Experimental results are provided to demonstrate the superior performance of the proposed approach.

No abstract

BackgroundGenetic, observational, and clinical intervention studies indicate that circulating levels of remnant cholesterol (RC) are associated with cardiovascular diseases. However, the predictive value of RC for cardiovascular mortality in the general population remains unclear.MethodsOur study population comprised 19,650 adults in the United States from the National Health and Nutrition Examination Survey (NHANES) (1999–2014). RC was calculated from non-high-density lipoprotein cholesterol (non-HDL-C) minus low-density lipoprotein cholesterol (LDL-C) determined by the Sampson formula. Multivariate Cox regression, restricted cubic spline analysis, and subgroup analysis were applied to explore the relationship of RC with cardiovascular mortality.ResultsThe mean age of the study cohort was 46.4 ± 19.2 years, and 48.7% of participants were male. During a median follow-up of 93 months, 382 (1.9%) cardiovascular deaths occurred. In a fully adjusted Cox regression model, log RC was significantly associated with cardiovascular mortality [hazard ratio (HR) 2.82; 95% confidence interval (CI) 1.17–6.81]. The restricted cubic spline curve indicated that log RC had a linear association with cardiovascular mortality (p for non-linearity = 0.899). People with higher LDL-C (≥130 mg/dL), higher RC [≥25.7/23.7 mg/dL in males/females corresponding to the LDL-C clinical cutoff point (130 mg/dL)] and abnormal HDL-C (<40/50 mg/dL in males/females) levels had a higher risk of cardiovascular mortality (HR 2.18; 95% CI 1.13–4.21 in males and HR 2.19; 95% CI 1.24–3.88 in females) than the reference group (lower LDL-C, lower RC and normal HDL-C levels).ConclusionsElevated RC levels were associated with cardiovascular mortality independent of traditional risk factors.

No abstract