bims-aukdir Biomed News
on Automated knowledge discovery in diabetes research
Issue of 2025–05–25
fourteen papers selected by
Mott Given



  1. Quant Imaging Med Surg. 2025 May 01. 15(5): 4816-4846
       Background and Objective: In the field of ophthalmology, diabetic retinopathy (DR) is a diabetes-related eye condition that damages the retina. DR is a serious issue for working-age people and it is important to solve it at an early stage to avoid complete vision loss. The review covers together technical and clinical perspectives of using the artificial intelligence (AI)-based systems in clinical institutes to help the ophthalmologists interpret and diagnose DR at the early stages.
    Methods: AI plays a significant role in assisting ophthalmologists with timely and effective treatment of patients by early and accurate detection and classification of DR. The study selection method followed specified searching criteria to complete the data collection task in the area of DR screening through AI.
    Key Content and Findings: The review covers literature published in the nearest one decade or more for analyzing automated DR diagnostics through the identification of retinal lesions and evaluates the approaches of advanced AI-based models for the development of early and accurate DR diagnostic processes. The DR-related datasets and performance evaluation metrics used for segmentation and classification tasks of DR, are also included in this review. Moreover, the authors performed critical analysis and provided possible solutions against DR-based potential problems.
    Conclusions: Hence, the study serves as a helping guide for the researchers to utilize their skills with advanced AI-based approaches in the technical and clinical perspectives of DR diagnostics. In conclusion, a freely available repository is created to facilitate the relevant researchers with up-to-date articles and open-source implementations for DR screening at https://github.com/muhammadmateen319/progress-of-diabetic-retinopathy-screening.
    Keywords:  Artificial intelligence (AI); diabetic retinopathy (DR); explainable AI (XAI); meta learning; multimodality
    DOI:  https://doi.org/10.21037/qims-24-1791
  2. PLoS One. 2025 ;20(5): e0324382
       INTRODUCTION: Delayed diagnosis of diabetic retinopathy (DR) remains a significant challenge, often leading to preventable blindness and visual impairment. Given that physicians are frequently the first point of contact for people with diabetes, there is a critical need for integrated screening programs within diabetes clinics to enhance DR management and reduce the risk of severe vision loss.
    METHODS AND ANALYSIS: We will conduct a prospective cohort study comparing (i) the intervention cohort, screened at diabetes clinics and referred to eye clinics per the proposed pathway, and (ii) the standard-of-care (SOC) eye clinic cohort. The study will be conducted in Hyderabad, India, at LV Prasad Eye Institute and four IDEA (Institute of Diabetes, Endocrinology, and Adiposity) Clinics. The primary objective is to evaluate the effectiveness of a systematic diabetic retinopathy screening program in achieving earlier detection and reducing visual impairment among People With Diabetes (PWD) attending IDEA clinics compared to routine care at eye care settings. The screening program will be operationalized using AI-enabled tools and supported by trained non-medical technicians. We will perform visual acuity tests and non-mydriatic fundus photography using AI-assisted cameras. DR-positive patients will be referred for treatment and follow-up. We aim to achieve high accuracy (>90%) in appropriate referral of DR and high screening coverage (>80%) of eligible PWD. Success metrics include screening uptake, AI diagnostic accuracy, referral rates, cost-effectiveness, patient satisfaction, follow-up adherence, and long-term outcomes.
    CONCLUSION: This study aims to enhance diabetic retinopathy screening and management through an AI-enabled approach at diabetes clinics, improving early detection and care pathways. The findings will contribute to evidence-based strategies for optimizing DR screening and management, with results disseminated through peer-reviewed publications to inform policy and practice.
    TRIAL REGISTRATION: Trial registration number: CTRI/2024/03/064518 [Registered on: 20/03/2024] (https://ctri.nic.in/).
    DOI:  https://doi.org/10.1371/journal.pone.0324382
  3. Stud Health Technol Inform. 2025 May 15. 327 1135-1139
      Optical Coherence Tomography (OCT) has become an indispensable imaging modality in ophthalmology, providing high-resolution cross-sectional images of the retina. Accurate classification of OCT images is crucial for diagnosing retinal diseases such as Age-related Macular Degeneration (AMD) and Diabetic Macular Edema (DME). This study explores the efficacy of various deep learning models, including convolutional neural networks (CNNs) and Vision Transformers (ViTs), in classifying OCT images. We also investigate the impact of integrating metadata (patient age, sex, eye laterality, and year) into the classification process, even when a significant portion of metadata is missing. Our results demonstrate that multimodal models leveraging both image and metadata inputs, such as the Multimodal ResNet18, can achieve competitive performance compared to image-only models, such as DenseNet121. Notably, DenseNet121 and Multimodal ResNet18 achieved the highest accuracy of 95.16%, with DenseNet121 showing a slightly higher F1-score of 0.9313. The multimodal ViT-based model also demonstrated promising results, achieving an accuracy of 93.22%, indicating the potential of Vision Transformers (ViTs) in medical image analysis, especially for handling complex multimodal data.
    Keywords:  Machine Learning; Multimodal Deep Learning; Optical Coherence Tomography; Retinal Disease Classification; Vision Transformers
    DOI:  https://doi.org/10.3233/SHTI250567
  4. Int J Med Inform. 2025 May 17. pii: S1386-5056(25)00192-3. [Epub ahead of print]202 105975
       PURPOSE: Machine learning (ML) has gained attention in diabetes management, particularly for predicting and diagnosing diabetic kidney disease (DKD). However, systematic evidence on its performance remains limited. This study evaluates the predictive and diagnostic accuracy of ML in DKD to support the development of tailored prevention strategies and non-invasive diagnostic tools.
    METHODS: A systematic search of PubMed, Embase, Web of Science, and Cochrane (up to April 14, 2024) identified relevant studies. Risk of bias was assessed using tools for predictive models, and meta-analysis included subgroup analyses based on task type, dataset, and model type.
    RESULTS: A total of 34 studies were included, with 19 on DKD risk prediction and 15 on diagnosis. For prediction, the pooled c-index was 0.81 (95% CI 0.79-0.83), sensitivity 0.81 (95% CI 0.74-0.86), and specificity 0.82 (95% CI 0.73-0.89). For diagnosis, the pooled c-index was 0.81 (95% CI 0.79-0.83), sensitivity 0.81 (95% CI 0.78-0.84), and specificity 0.75 (95% CI 0.72-0.79).
    CONCLUSIONS: ML shows promising accuracy in DKD prediction and diagnosis, offering a viable tool for early screening and risk assessment.
    Keywords:  Diabetic kidney disease; Diagnosis; Machine learning; Meta-analysis; Prediction
    DOI:  https://doi.org/10.1016/j.ijmedinf.2025.105975
  5. Graefes Arch Clin Exp Ophthalmol. 2025 May 16.
      Despite current screening models, enhanced imaging modalities, and treatment regimens, diabetic retinopathy (DR) remains one of the leading causes of vision loss in working age adults. DR can result in irreversible structural and functional retinal damage, leading to visual impairment and reduced quality of life. Given potentially irreversible photoreceptor damage, diagnosis and treatment at the earliest stages will provide the best opportunity to avoid visual disturbances or retinopathy progression. We will review herein the current structural imaging methods used for DR assessment and their capability of detecting DR in the first stages of disease. Imaging tools, such as fundus photography, optical coherence tomography, fundus fluorescein angiography, optical coherence tomography angiography and adaptive optics-assisted imaging will be reviewed. Finally, we describe the future of DR screening programmes and the introduction of artificial intelligence as an innovative approach to detecting subtle changes in the diabetic retina. CLINICAL TRIAL REGISTRATION NUMBER: N/A.
    Keywords:  Detection; Diabetes; Diabetic retinopathy; Fundus photography; Imaging; Optical coherence tomography
    DOI:  https://doi.org/10.1007/s00417-025-06828-3
  6. J Imaging Inform Med. 2025 May 20.
      Generative adversarial networks (GANs), introduced by Ian Goodfellow in 2014, have revolutionized adversarial machine learning, particularly in data synthesis. This manuscript explores their application in ophthalmic diagnostics, addressing the scarcity of annotated datasets and the need for improved early disease detection. By leveraging GAN architectures, the goal is to enhance the quality of synthetic ophthalmic images, ultimately improving diagnostic algorithm training. A systematic review was conducted from January to April 2024 across PubMed, Embase, and Scopus. Search terms included "Generative Adversarial Networks" and "ophthalmic image synthesis." Articles were selected based on relevance to retinal image generation and diagnostic improvement in ophthalmology. GANs show considerable promise in generating high-resolution retinal and optical coherence tomography (OCT) images. Models like DR-GAN and Pix2Pix have successfully synthesized images that resemble real diagnostic data, proving valuable when annotated datasets are scarce. GAN-generated images enhance training for algorithms detecting diseases such as diabetic retinopathy, glaucoma, and age-related macular degeneration. Recent advances, including conditional GANs and CycleGANs, have enabled disease-specific image generation, boosting the diversity of training datasets, particularly in resource-limited settings. Integrating GANs into ophthalmic diagnostics represents a significant leap in medical AI, offering high-quality synthetic images to improve diagnostic algorithms. Despite their potential, challenges such as the need for larger datasets, improved image interpretability, and noise reduction must be addressed. Future research should focus on optimizing these models and incorporating multi-modal data to enhance diagnostic accuracy in clinical settings.
    Keywords:  Diffusion models; GAN; Generative adversarial networks; Image synthesis; OCT
    DOI:  https://doi.org/10.1007/s10278-025-01519-1
  7. Comput Methods Programs Biomed. 2025 May 12. pii: S0169-2607(25)00229-9. [Epub ahead of print]268 108812
       BACKGROUND AND OBJECTIVE: Diabetes is a chronic disease characterised by a high risk of developing diabetic nephropathy. The early identification of individuals at heightened risk of such complications or their exacerbation can be crucial to set a correct course of treatment. However, there are currently no widely accepted predictive tools for this task and, additionally, most of these models rely only on information at a single baseline visit. Considering this, we investigate the potential predictive role of patients' clinical history over multiple levels of renal disease severity while, at the same time, developing an effective predictive model.
    METHODS: From the data collected in the DARWIN-Renal (DApagliflozin Real-World evIdeNce-Renal) study, a nationwide multicentre retrospective real-world study, we develop four different types of machine learning models, namely, logistic regression, random forest, Cox proportional hazards regression, and a deep learning model based on recurrent neural network to predict the crossing of 5 clinically relevant glomerular filtration rate thresholds for patients with type 2 diabetes.
    RESULTS: The predictive performance of all models is satisfactory for all outcomes, even without the introduction of information referring to past visits, with AUROC and C-index between 0.69 and 0.98 and average precision well above the random model. The introduction of past information results into a clear improvement in performance for all the models, with percentage increases of up to 12% for both AUROC and C-index and 300% for average precision. The usefulness of past information is further corroborated by a feature importance analysis.
    CONCLUSIONS: Incorporating data from the patients' clinical history into the predictive models greatly improves their performance, particularly for recurrent neural network where the full sequence of values for dynamic variables is provided compared to synthetic indicators of past history.
    Keywords:  Deep learning; Diabetes; Kidney disease; Longitudinal data; Machine learning; Predictive modelling
    DOI:  https://doi.org/10.1016/j.cmpb.2025.108812
  8. Cureus. 2025 Apr;17(4): e82437
       INTRODUCTION: Hospital readmissions within 30 days remain a critical issue in healthcare, signaling potential care discontinuities and contributing to escalating costs. Leveraging machine learning (ML) on electronic health records (EHRs) presents a promising strategy to identify patients at heightened risk for readmission and support early intervention.
    AIM: This study evaluates the performance of four ML models - logistic regression, random forest, XGBoost, and deep neural networks (DNN) - in predicting 30-day hospital readmissions. It also identifies the most influential predictors using SHapley Additive exPlanations (SHAP) values.
    MATERIALS AND METHODS: We conducted a retrospective analysis on 101,766 de-identified inpatient encounters from the University of California, Irvine (UCI) Diabetes 130-United States (US) Hospitals dataset. After preprocessing, including feature imputation, scaling, and one-hot encoding, we trained and validated models on an 80/20 train-test split. Evaluation metrics included accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (AUC-ROC).
    RESULTS: XGBoost achieved the highest AUC-ROC (0.667), followed closely by logistic regression (0.642) and random forest (0.630). Despite DNNs demonstrating the highest recall for the positive class (0.143), their AUC-ROC (0.579) and precision (0.186) indicated lower reliability. SHAP analysis revealed that previous admissions, number of medications, and comorbidity indicators such as diabetes medication usage and admission type were key predictors influencing model decisions.
    CONCLUSION: XGBoost outperformed other models in predicting 30-day readmissions using EHR data, balancing performance and interpretability when coupled with SHAP values. These findings underscore the promise of ensemble models in improving discharge planning and reducing preventable readmissions. Future research should explore the inclusion of behavioral and social health features to further enhance predictive accuracy.
    Keywords:  diabetes dataset; ehr; healthcare analytics; hospital readmission; logistic regression model; machine learning; shap; xgboost
    DOI:  https://doi.org/10.7759/cureus.82437
  9. Diabetol Metab Syndr. 2025 May 17. 17(1): 159
      Cardiovascular disease complications are the leading cause of morbidity and mortality in patients with Type 2 diabetes (T2DM). Left ventricular diastolic dysfunction (LVDD) is one of the earliest myocardial characteristics of diabetic cardiac dysfunction. Therefore, we aimed to develop an LVDD-risk predictive model to diagnose cardiac dysfunction before severe cardiovascular complications arise. We trained an artificial neural network model to predict LVDD risk with patients' clinical information. The model showed better performance than classical machine learning methods such as logistic regression, random forest and support vector machine. We further explored LVDD-risk/protective features with interpretability methods in neural network. Finally, we provided a freely accessible web server called LVDD-risk, where users can submit their clinical information to obtain their LVDD-risk probability and the most noteworthy risk indicators.
    Keywords:  Cardiovascular disease; Diabetic complications; Left ventricular diastolic dysfunction; Machine learning; Neural network
    DOI:  https://doi.org/10.1186/s13098-025-01714-8
  10. BMC Oral Health. 2025 May 22. 25(1): 765
       BACKGROUND: Diabetic oral ulceration (DOU) is a prevalent and debilitating complication among diabetic patients, significantly impairing their quality of life and imposing substantial economic burdens. Studies indicate that over 90% of diabetic patients experience oral complications, with 45% suffering from oral ulcers. Clear diagnosis is crucial for effective clinical management and prognosis improvement. However, current diagnostic methods often fall short in early detection and intervention. Machine learning (ML) has shown promise in predicting disease development, yet no relevant predictive models for DOU have been established.
    METHODS: This study aimed to develop an ML-based predictive model for DOU using oral examination, clinical, and socioeconomic data. The dataset included 324 diabetic patients, with 127 DOU features. One-hundred-fold cross-validation was employed for model optimization and feature selection. Data preprocessing involved handling missing values, scaling different range values, and feature selection using techniques such as Variance Threshold (VT), Mutual Information (MI), and Variance Inflation Factor (VIF). Four prediction models, Support Vector Machine Classifier (SVC), Multi-layer Perceptron (MLP), Logistic Regression Classifier (LogReg), and Perceptron, were established and evaluated.
    RESULTS: The SVC model outperformed the other models, achieving an accuracy (ACC) of 0.95 and an area under the ROC curve (AUC) of 0.91. The top five features contributing to the model's predictions were the current number of oral ulcers, diminished oral functional capacity, number of decayed or missing teeth, possession of health insurance (commercial), and Low-Density Lipoprotein (LDL-C), accounting for 57.32% of the total importance. Oral examination indicators accounted for 46.46%, serum lipid markers for 6.93%, and sociodemographic factors, personal lifestyles, and cardiovascular diseases also played significant roles.
    CONCLUSION: The SVC model demonstrated superior performance and stability, making it suitable for predicting DOU occurrence and development in diabetic patients. This study's innovation lies in the comprehensive evaluation of multiple factors, including oral examinations, physiological indicators, self-management capabilities, and economic factors, to facilitate efficient DOU screening. The findings highlight the potential of ML in improving diagnostic accuracy and enabling timely interventions for DOU, ultimately contributing to better clinical management and patient outcomes. Future research should focus on validating the model across larger, multicenter cohorts and further exploring the long-term impact of ML-guided interventions on DOU management.
    Keywords:  Diabetes; Machine learning; Oral ulcer; Predictive model
    DOI:  https://doi.org/10.1186/s12903-025-06096-x
  11. Clin Lab. 2025 May 01. 71(5):
       BACKGROUND: Epstein-Barr virus (EBV) is a ubiquitous herpesvirus that is known to cause infectious mononucleosis and is associated with several autoimmune diseases and cancers through immune system dysregulation and chronic inflammatory mechanisms.
    METHODS: The authors collected 3,624 samples containing EBV DNA test results and 1,872 samples containing EBV antibody test results from Qingdao Central Hospital. The machine learning model was trained using CatBoost classifier, and the data imbalance problem was dealt with using SMOTE method. For the EBV antibody data, normality was assessed using the Shapiro-Wilk test, and the Welch's t-test and Mann-Whitney U test were used to compare the differences between the type 2 diabetic and non-diabetic groups. Finally, the causal relationship between EBV antibodies and type 2 diabetes was verified by Mendelian randomization.
    RESULTS: Machine learning modeling showed 70% prediction accuracy of EBV DNA in immunoendocrine diseases. Type 2 diabetic patients had significantly higher VCA IgG levels than non-diabetic patients (p < 0.05). Mendelian randomization analysis further validated the positive correlation between type 2 diabetes mellitus and VCA IgG levels (p < 0.05), suggesting that patients with type 2 diabetes mellitus may have higher VCA IgG levels.
    CONCLUSIONS: This study found a significant association between type 2 diabetes and EBV VCA IgG levels, emphasizing the potential relationship between EBV infection and diabetes. Machine learning and Mendelian randomization methods played an important role in determining disease associations, which provides new ideas for future clinical management and prevention strategies.
    DOI:  https://doi.org/10.7754/Clin.Lab.2025.250137