Madras Diabetes Research Foundation

Diabetic Retinopathy (DR) is a serious microvascular complication of diabetes, and one of the leading causes of vision loss worldwide. Although automated detection and grading, with Deep Learning (DL), can reduce the burden on ophthalmologists, it is constrained by the limited availability of high-quality datasets. Existing repositories often remain geographically narrow, contain limited samples, and exhibit inconsistent annotations or variable image quality; thereby, restricting their clinical reliability. This paper presents a comprehensive review and comparative analysis of fundus image datasets used in the management of DR. The study evaluates their usability across key tasks, including binary classification, severity grading, lesion localization, and multi-disease screening. It also categorizes the datasets by size, accessibility, and annotation type (such as image-level, lesion-level, and multi-disease). Finally, a recently published dataset is presented as a case study to illustrate broader challenges in dataset curation and usage. The review consolidates current knowledge while highlighting persistent gaps such as the lack of standardized lesion-level annotations and longitudinal data. It also outlines recommendations for future dataset development to support clinically reliable and explainable solutions in DR screening.

Abstract Background Glycated haemoglobin (HbA1c) underpins type 2 diabetes (T2D) and prediabetes management worldwide and reflects both glycaemia and erythrocyte biology. A missense variant in PIEZO1 (rs563555492 T ), carried by 1 in 12 south Asians, has been associated with a nonglycemic reduction in HbA1c. We aimed to further characterise this association and evaluate its clinical consequences. Methods We undertook genetic and linked health data analyses across two cohorts: 19,898 (37.4% female) South Indians from the Madras Diabetes Research Foundation (MDRF) and 43,011 (54.4% female) British Bangladeshis and British Pakistanis in Genes & Health. In MDRF, we tested associations with glycaemic and erythrocytic traits using additive genetic models. In Genes & Health we modelled diagnosis of prediabetes, T2D, and diabetic eye disease using flexible parametric survival models. Ten-year absolute risks were estimated for a population aged 40-50 years. Findings PIEZO1 rs563555492 T was associated with erythrocytic traits and lower HbA1c, but not with fasting glucose, postprandial glucose, or C-peptide. This variant reduced risk of prediabetes (HR 0·63, 95% CI 0·58-0·69) and T2D (0·85, 0·78-0·93) diagnosis, and increased risk of diabetic eye disease among individuals with T2D (1·20, 1·01-1·43). Modelling suggested approximately 1,019 missed prediabetes and 303 missed T2D diagnoses per 100,000 adults over 10 years. Interpretation An ancestry enriched PIEZO1 variant is associated with lower HbA1c independent of glycaemia, reduced prediabetes and T2D diagnosis suggesting delayed detection, and increased complication risk. Reliance on HbA1c may systematically underestimate glycaemic risk in a substantial minority of south Asians. Funding The Wellcome Trust; NIHR Research in Context Evidence before this study We searched PubMed from database inception to Nov 1, 2025, for studies examining PIEZO1, Glycated Haemoglobin (HbA1c) and type 2 diabetes (T2D) related outcomes. Search terms included “PIEZO1”, “haemoly*”, “hemol*”, “erythrocyte*”, “HbA1c”, “glycated haemoglobin”, “glycated hemoglobin”, “type 2 diabetes”, “prediabetes”, “retinopathy”, and related terms. HbA1c is widely used to guide prevention, diagnosis, and management of T2D, but variants affecting erythrocyte structure or lifespan can alter its relationship with glycaemia. A missense variant in PIEZO1 (rs563555492 T ), carried by one in twelve south Asians, has been associated with lower HbA1c and erythrocytic biomarkers without differences in fasting glucose and with older age at T2D diagnosis, suggesting delayed diagnosis in UK-based cohorts. However, uncertainties remain. Associations have not been replicated outside the UK to confirm generalisability. Experimental animal models suggest PIEZO1 channels may influence pancreatic insulin secretion, but this has not been evaluated in human cohorts. Finally, the potential clinical and health-economic consequences associated with this variant remain largely unexplored. Added value of this study We replicated associations between rs563555492 T , lower HbA1c, and erythrocytic changes in a geographically distinct south Asian cohort based in South India. We found no association with fasting glucose, postprandial glucose, or C-peptide concentrations, supporting a predominantly erythrocytic mechanism. We examined the clinical and economic consequences of this variant on south Asians in a large UK population-based cohort, following introduction of HbA1c-based diagnostic criteria. Carriers experienced delayed diagnosis of prediabetes and T2D, and there was some evidence of increased risk of diabetic eye disease. In a modelled population aged 40-50 years, the variant was associated with approximately 1,019 missed prediabetes diagnoses and 303 missed T2D diagnoses per 100,000 adults over 10 years. Because prediabetes diagnosis initiates referral to prevention programmes, missed diagnoses represent lost opportunities for disease prevention. In a UK NHS setting, we estimated the associated opportunity cost to be between £970,000 - £1,390,000 per 100,000 individuals over a 10-year horizon. Implications of all the available evidence A south Asian ancestry enriched PIEZO1 variant can disrupt the relationship between HbA1c and glycaemia and delay detection of prediabetes and T2D. Reliance on HbA1c alone may underestimate glycaemic risk in a substantial minority of individuals of south Asian ancestry, with potential implications for prevention strategies, risk of complications and health care costs. Alternative biomarkers of glycaemia or individualisation of HbA1c may help ensure timely and accurate diabetes detection in populations where this variant is common.

OBJECTIVE To investigate associations between polyendocrine metabolic ovarian syndrome (PMOS) and maternal and neonatal outcomes and explore interactions with BMI and ethnicity. RESEARCH DESIGN AND METHODS This was a prospective nested cohort study of pregnant women with risk factors for hyperglycemia enrolled in an international, multicenter, randomized controlled trial of gestational diabetes mellitus (GDM) treatment. We compared baseline and pregnancy characteristics, evaluated adverse maternal and neonatal outcomes by PMOS status, and used multivariable regression models to evaluate factors (maternal age, baseline BMI, ethnicity, parity, smoking, and level of education) that independently affected maternal and neonatal outcomes. RESULTS Of 3,645 participants, PMOS prevalence was 17.1% (95% CI 15.9, 18.3). At booking visits, women with PCOS (vs. without) were younger (30.6 ± 4.7 vs. 31.3 ± 5.2 years; P < 0.001) and had higher median (interquartile range) BMI at baseline (29.8 [25.1–35.7] vs. 28.1 [24.1–33.9] kg/m2; P < 0.001), and more were primigravid (30.2% [n = 188] vs. 25.7% [n = 778]; P = 0.022). There were positive associations between PCOS and early GDM, with an adjusted odds ratio (aOR) of 1.37 (95% CI 1.10, 1.72), a composite of adverse neonatal outcomes (aOR 1.28 [95% CI 1.04, 1.58]), admission to a neonatal special care nursery or neonatal intensive care unit (aOR 1.32 [95% CI 1.05–1.65]), and negative associations between PMOS and gestational age at birth (−1.60 days [95% CI −2.82, −0.39]) and birth length (−0.33 cm [95% CI −0.61, −0.04]). No interactions were found with baseline BMI and ethnicity. CONCLUSIONS PMOS was associated with higher odds of early GDM, and neonatal outcomes were poorer on adjusted analyses, highlighting independent pregnancy risks in PMOS and the need to identify, monitor, and treat women with PMOS to mitigate risks of adverse pregnancy outcomes.

Dementia continues to pose a growing challenge worldwide, particularly in aging populations where effective disease-modifying treatments remain limited. In this context, attention has increasingly shifted toward non-pharmacological approaches, including physical activity and structured exercise. Evidence accumulated over the past decade suggests that regular exercise may offer modest but meaningful benefits across several domains. These include improvements in cognitive function—especially executive function and attention—as well as gains in mobility, daily functioning, and psychological well-being. While the extent of cognitive improvement varies across studies, a consistent pattern emerges indicating greater benefit when interventions are initiated earlier in the disease course. Beyond clinical outcomes, several biological mechanisms have been proposed, including enhanced cerebral perfusion, upregulation of neurotrophic factors such as brain-derived neurotrophic factor, and attenuation of neuroinflammatory processes. At the same time, challenges remain. Differences in exercise protocols, issues related to adherence, and variability in patient populations make it difficult to define an optimal approach. Despite these limitations, physical activity represents a practical, low-cost strategy that can be integrated into routine care. Further research is needed to refine intervention design and understand long-term effects.