bims-aukdir Biomed News
on Automated knowledge discovery in diabetes research
Issue of 2026–02–22
twenty papers selected by
Mott Given
  1. Graph-enhanced multimodal fusion of vascular biomarkers and deep features for diabetic retinopathy detection.
  2. A machine learning approach to predicting severe diabetes distress in adults with type 1 diabetes mellitus.
  3. Comparison of Interpretable Machine Learning Models Using Systemic Inflammation Index to Predict Preterm Birth in Gestational Diabetes Mellitus.
  4. Hypoglycemia problem-solving ability as a key predictor of severe hypoglycemia: a machine learning approach in adults with type 1 diabetes.
  5. Corrigendum to "Predicting postpartum glucose intolerance in women with gestational diabetes mellitus in primary care: A machine learning approach using XGBoost and SHAP values" [Diabetes Res. Clin. Pract. 232 (2026) 113098].
  6. OPU-NET-DADENA: Optimized deep learning ensemble with u-net segmentation for early detection of diabetic retinopathy.
  7. Interpretable machine learning for predicting in-hospital bleeding after PCI in patients with diabetes and acute coronary syndrome: a retrospective cohort study.
  8. Application of machine learning approaches to develop predictive models for diabetes and hypertension among Bangladeshi Adults.
  9. Deep Learning-Enabled Diabetic Retinopathy Screening: A Techno-Clinical Revolution or Just More Artificial Intelligence Hype?
  10. Application study of an artificial intelligence and big data-based personalized chronic disease management model for diabetes patients.
  11. Smart Kiosk for Nutritional Management of People With Diabetes in Underserved Communities: Development and Technical Evaluation.
  12. Algor-Ethics in Diabetes Care: Mapping the Route.
  13. Exploring the feasibility of near-infrared spectroscopy and machine learning for detecting cardiovascular diseases and diabetes mellitus in fingernails.
  14. Evaluation of validity, reliability, and readability of AI chatbots for gestational diabetes mellitus: a multi-model comparative study.
  15. A machine learning model for optimizing treatment of patients with poorly controlled type 2 diabetes.
  16. Re: Henriques et al Correspondence Re: Most et al: Can Multimodal Large Language Models Diagnose Diabetic Retinopathy from Fundus Photos? A Quantitative Evaluation.
  17. A Hybrid Ensemble Approach for Early-Stage Diabetes Detection.
  18. Re: Most et al: Can Multimodal Large Language Models Diagnose Diabetic Retinopathy from Fundus Photos? A Quantitative Evaluation.
  19. Identifying Patients at High Risk of Lapses in Diabetic Retinopathy Care: A Machine Learning Study.
  20. Deep-Q-Network in Multiview Ensemble Learning to Predict Anti-diabetic Peptide and Diabetic Types.
  1. Front Artif Intell. 2025 ;8 1731633
    Graph-enhanced multimodal fusion of vascular biomarkers and deep features for diabetic retinopathy detection.
    K V Deepsahith, Basineni Shashank, Bangipavan Kumar, Sherly Alphonse, Brindha Subburaj, Girish Subramanian.
      Diabetic retinopathy (DR) detection can be performed through both deep retinal representations and vascular biomarkers. This proposed work suggests a multimodal framework that combines deep features with vascular descriptors in transformer fusion architecture. Fundus images are preprocessed using CLAHE, Canny edge detection, Top-hat transformation, and U-Net vessel segmentation. Then, the images are passed through a convolutional block attention module (CBAM)-fused enhanced MobileNetV3 backbone for deep spatial feature extraction. In parallel, the segmented vasculature is skeletonized to create a vascular graph, and the descriptors are computed using fractal dimension analysis (FDA), artery-to-vein ratio (AVR), and gray level co-occurrence matrix (GLCM) texture. A graph neural network (GNN) then generates a global topology-aware embedding using all that information. The different modalities are integrated using a transformer-based cross-modal fusion, where the feature vectors from MobileNet and GNN-based vascular embeddings interact using multi-head cross-attention. The fused representation is then given to a Softmax classifier for DR prediction. The model demonstrates superior performance compared to traditional deep learning baselines, achieving an accuracy of 93.8%, a precision of 92.1%, a recall of 92.8%, and an AUC-ROC of 0.96 for the DR prediction in the Messidor-2 dataset. The proposed approach also achieves above 98% accuracy for Eyepacs and APTOS 2019 datasets for DR detection. The findings demonstrate that the proposed system provides a reliable framework compared with the existing state-of-the-art methods.
    Keywords:  Convolutional Neural Networks (CNNs); MobileNetV3; contrast limited adaptive histogram equalization (CLAHE); deep learning; retinal images
    DOI:  https://doi.org/10.3389/frai.2025.1731633
  2. J Diabetes Investig. 2026 Feb 18.
    A machine learning approach to predicting severe diabetes distress in adults with type 1 diabetes mellitus.
    Naoki Sakane, Yushi Hirota, Akane Yamamoto, Junnosuke Miura, Hiroko Takaike, Sari Hoshina, Masao Toyoda, Nobumichi Saito, Kiminori Hosoda, Masaki Matsubara, Atsuhito Tone, Satoshi Kawashima, Hideaki Sawaki, Tomokazu Matsuda, Masayuki Domichi, Akiko Suganuma, Seiko Sakane, Takashi Murata.
       BACKGROUND: Diabetes distress is common in patients with type 1 diabetes mellitus (T1DM). The aim of this study was to construct and validate prediction models for diabetes distress in adults with T1DM using continuous glucose monitoring (CGM) metrics.
    METHODS: The CGM metrics were collected from 259 adults with T1DM. Severe diabetes distress was defined as 40 points on the Problem Areas in Diabetes scale. Prediction models were developed based on ten machine learning algorithms: random forest (RF), support vector machine (SVM), Naive Bayes (NB), Neural Network (NN), k-nearest neighbor (k-NN), XGBoost (XGB), SGDClassifier (SGDC), XGB_limitet_depth (CGB_ld), L1LogisticRegression (L1), and LightGBM. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, precision, recall, and F1 score.
    RESULTS: Among the ten models, accuracy in the NN model was the highest (NN: 0.744, L1: 0.731, NB: 0.718, SVM: 0.692, SGDC: 0.654, RF: 0.628, k-NN: 0.628, LightGBM: 0.615, XGBoost: 0.564, and XGB_ld: 0.564). The NN model achieved the highest AUC of 0.728 (95% confidence interval: 0.608-0.845).
    CONCLUSIONS: This study developed a predictive model for severe diabetes distress using machine learning, incorporating both demographic and CGM metrics in adults with type 1 diabetes mellitus. The NN model demonstrated potential as a practical tool to assist clinicians in identifying individuals at risk of severe diabetes distress.
    Keywords:  Machine learning; Neural Network; type 1 diabetes mellitus
    DOI:  https://doi.org/10.1111/jdi.70229
  3. Int J Womens Health. 2026 ;18 541610
    Comparison of Interpretable Machine Learning Models Using Systemic Inflammation Index to Predict Preterm Birth in Gestational Diabetes Mellitus.
    Qinxia Pang, Lei Peng, Jianfa Wu, Ying Wang, Rong Zhang, Zhou Liu, Lingli Jiang.
       Background: Gestational diabetes mellitus (GDM) elevates preterm birth risk, highlighting the need for improved prediction methods to enhance outcomes. Current models show limited accuracy by ignoring some inflammatory biomarkers (eg, PLR, LMR, SII). Machine learning (ML) can better analyze complex patterns but remains underused for GDM preterm birth prediction.
    Objective: This study develops an interpretable ML model combining systemic inflammatory indices and traditional clinical markers to predict preterm birth in GDM. Enabling early risk stratification at diagnosis, it facilitates timely interventions for this high-risk population.
    Methods: This retrospective study analyzed 389 GDM patients, stratified into training (n=272) and temporal external validation (n=117) cohorts, and further classified by birth outcome (term/preterm). Using the training cohort, we developed and internally validated multiple ML models incorporating: (1) systemic inflammation indices, (2) traditional clinical indicators, and (3) their combination. The optimal model underwent temporal external validation and subsequent Shapley Additive Explanations (SHAP) analysis for feature interpretation. To assess the robustness of our findings, sensitivity analyses were conducted.
    Results: Our cohort of 389 GDM patients included 53 preterm births (13.6%). Analysis revealed seven significant predictors combining systemic inflammatory markers and traditional clinical parameters. The extreme gradient boosting (XGBoost) model outperformed comparative algorithms (AUC-ROC: 0.932 vs Logit: 0.871, SVM: 0.847, RF: 0.917; AUC-PRC: 0.754 vs Logit: 0.686, SVM: 0.582, RF: 0.670). SHAP analysis identified five key determinants (two clinical and three inflammatory markers) as most influential for preterm birth prediction. Sensitivity analyses were conducted to assess the robustness of the results.
    Conclusion: The XGBoost model outperforms in predicting GDM-related preterm birth by integrating traditional clinical and systemic inflammatory markers, enabling precise risk assessment to guide clinical management.
    Keywords:  gestational diabetes mellitus; machine learning; preterm birth; systemic inflammation index
    DOI:  https://doi.org/10.2147/IJWH.S541610
  4. Diabetol Int. 2026 Apr;17(2): 24
    Hypoglycemia problem-solving ability as a key predictor of severe hypoglycemia: a machine learning approach in adults with type 1 diabetes.
    Naoki Sakane, Ken Kato, Sonyun Hata, Erika Nishimura, Rika Araki, Kunichi Kouyama, Masako Hatao, Yuka Matoba, Yuichi Matsushita, Masayuki Domichi, Akiko Suganuma, Seiko Sakane, Takashi Murata, Fei Ling Wu.
       Aims/introduction: Severe hypoglycemia (SH) is a major complication in adults with type 1 diabetes mellitus (T1DM). The multifactorial etiology of T1DM highlights the need for predictive tools that integrate clinical, behavioral, and technological factors. This study aimed to develop and evaluate machine learning (ML) models for predicting SH by incorporating hypoglycemia problem-solving ability, diabetes technology, and continuous glucose monitoring (CGM) indices.
    Materials and methods: We analyzed data from 247 adults with T1DM (mean age 50.4 ± 13.7 years; 38.1% male; glycosylated hemoglobin 7.7 ± 0.9%) from the FGM-Japan study. A total of 22,517 feature-model combinations were evaluated across 11 ML algorithms, including logistic regression, L1-regularized regression, random forest, LightGBM, XGBoost, SVM, Naïve Bayes, SGD, neural networks, and k-nearest neighbors. Eleven candidate predictors included impaired awareness of hypoglycemia (IAH), diabetic peripheral neuropathy (DPN), CSII, rtCGM, and seven domains of hypoglycemia problem-solving ability. The model performance was assessed with fivefold cross-validation using the receiver operating characteristic-area under the curve (ROC-AUC), accuracy, precision, recall, and F1 score. Class imbalance was addressed using SMOTE.
    Results: The mean ROC-AUC across models was 0.64 (range: 0.151-0.916). The average accuracy was 0.90, but the precision and recall were consistently low, with a mean recall of 0.08. The high-performing models (ROC-AUC > 0.90) were primarily Random Forest and LightGBM, which frequently incorporated domains such as problem perception, identifying problem attributes, seeking preventive strategies, evaluating strategies, and immediate management. factors. Tree-based models significantly outperformed logistic regression, Naïve Bayes, SVM, and SGD (adjusted p < 0.001), whereas the differences among the tree-based algorithms were not clinically meaningful.
    Conclusions: Tree-based ML models demonstrated superior discriminative ability for predicting SH in patients with T1DM. Hypoglycemia problem-solving ability was the strongest predictor, underscoring the importance of integrating behavioral self-management skills with clinical and technological factors.
    Trial registration: University hospital Medical Information Network (UMIN) Center: UMIN000039475), Approval date 13 February 2020.
    Keywords:  Hypoglycemia problem-solving; Impaired awareness of hypoglycemia; Severe hypoglycemia; Type 1 diabetes
    DOI:  https://doi.org/10.1007/s13340-026-00875-9
  5. Diabetes Res Clin Pract. 2026 Feb 15. pii: S0168-8227(26)00073-2. [Epub ahead of print] 113154
    Corrigendum to "Predicting postpartum glucose intolerance in women with gestational diabetes mellitus in primary care: A machine learning approach using XGBoost and SHAP values" [Diabetes Res. Clin. Pract. 232 (2026) 113098].
    Nur Aiza Idris, Nurul Azreen Yusof, Muhd Zulfadli Hafiz Ismail, Siti Norazlina Juhari, Nurulhuda Mat Hassan, Norwati Daud, Nurul Izza Yunus, Mohd Faeiz Pauzi.
      
    DOI:  https://doi.org/10.1016/j.diabres.2026.113154
  6. Microvasc Res. 2026 Feb 18. pii: S0026-2862(26)00023-3. [Epub ahead of print] 104923
    OPU-NET-DADENA: Optimized deep learning ensemble with u-net segmentation for early detection of diabetic retinopathy.
    N Kavitha, S Sridevi.
      One of the most significant microvascular complications of diabetes mellitus (DM) is diabetic retinopathy (DR). In the early stages, patients with DR may not exhibit any noticeable symptoms. It is a diabetic disorder that damages retinal blood vessels in the eyes. At first, there are no symptoms or sporadic visual issues. When it worsens, it affects both eyes and can lead to partial or total blindness. A person who already has DM is therefore constantly at a higher risk of developing the condition. Early identification can prevent the possibility of total and irreversible blindness. Therefore, needs an efficient and early diagnosis system. So, this paper proposes a new deep-learning methodology in a specific Deep Siamese DenseNet for early detection of DR. The proposed method is accomplished through various steps, such as Data Collection, Preprocessing, Augmentation, Segmentation, Feature Extraction, Feature Selection, and Detection. An adaptive histogram equalization approach named Contrast Limited Adaptive Histogram Equalization (CLAHE) is used to preprocess input images, which reduces amplification of noise. Then, the segmentation is done by the Optimized U-NETs. Next, features of the contrasted retinal images are extracted by residual attention EfficientNet (RA-EfficientNet). Then, the optimal features are selected by a hybrid Reptile Search Algorithm (RORS). Finally, the deep learning methodology includes DarkNet, DenseNet 201, and NasNetMobile used to detect diabetic retinopathy at an early stage. The model is implemented in MATLAB and evaluated using accuracy, precision, F-score, specificity, sensitivity, MCC, NPV, FPR, and FNR. The proposed approach achieves 99.33% accuracy and 98.32% precision, outperforming previous methods that reported accuracies of 95-97% and precisions of 94-96%, demonstrating its effectiveness for reliable early detection of DR.
    Keywords:  Deep-learning; Diabetic retinopathy; Microvascular complications; Retinal blood vessels; Vision problems
    DOI:  https://doi.org/10.1016/j.mvr.2026.104923
  7. BMC Cardiovasc Disord. 2026 Feb 20.
    Interpretable machine learning for predicting in-hospital bleeding after PCI in patients with diabetes and acute coronary syndrome: a retrospective cohort study.
    Huasheng Lv, Ruotong Cao, Yuchen Zhang, Fengyu Sun, Yitong Ma, Xinrong Zhou.
      
    Keywords:  Acute Coronary Syndrome; Bleeding; Diabetes Mellitus; Machine Learning; Percutaneous Coronary Intervention; Prediction Model
    DOI:  https://doi.org/10.1186/s12872-026-05644-9
  8. PLOS Glob Public Health. 2026 ;6(2): e0004797
    Application of machine learning approaches to develop predictive models for diabetes and hypertension among Bangladeshi Adults.
    Gulam Muhammed Al Kibria, James Ross O'Hagan, Golam Shariar, Tarina Khan, Mohammed Elfaramawi.
      With rapid urbanization, lifestyle changes, and an aging population, non-communicable diseases (NCDs), including hypertension and diabetes, pose significant public health challenges in Bangladesh and many other low- and middle-income countries. This study used machine learning (ML) approaches to develop predictive models for hypertension and diabetes among Bangladeshi adults. Bangladesh Demographic and Health Survey 2022, a nationally representative cross-sectional survey, data were analyzed. Hypertension was defined as systolic/diastolic blood pressure 140/90 mmHg (or more) or taking any antihypertensive medication. Diabetes was defined as having fasting plasma glucose ≥7.0 mmol/L or using any glucose-lowering drugs. Potential predictors included age, sex, education, wealth quintile, overweight/obesity, rural-urban residence, and division of residence. Descriptive analysis was conducted, and six ML models were applied: artificial neural network (ANN), random forest, adaptive boosting (AdaBoost), gradient boosting, XGBoost, and support vector machine (SVM). Models' performance and feature importance were reported. We included 13,847 adults (females: 55%). Sensitivity was high across models (up to 0.96 and 0.90 for diabetes and hypertension, respectively). However, the overall specificity was low, particularly for diabetes. The prevalence of diabetes and hypertension was 16.3% and 20.5%, respectively. For diabetes, AdaBoost had the highest AUC (0.699), and SVM had the highest accuracy (0.836); for hypertension, AdaBoost had the greatest AUC (0.775) and accuracy (0.799). Hypertension was the most common diabetes predictor, while overweight/obesity was the most common predictor for hypertension, followed by age and diabetes. Wealth and sex were moderately influential, with education and geographic factors less so. Low specificity across models indicated challenges in identifying non-cases. This ML-driven analysis identified the bidirectional relationship of hypertension and diabetes along with several other predictors, including overweight/obesity, older age, and richer household wealth quintiles. Our findings underscore the need for integrated screening and lifestyle interventions targeting high-risk groups to mitigate future NCD burden.
    DOI:  https://doi.org/10.1371/journal.pgph.0004797
  9. Diabetes Care. 2026 Mar 01. 49(3): 381-383
    Deep Learning-Enabled Diabetic Retinopathy Screening: A Techno-Clinical Revolution or Just More Artificial Intelligence Hype?
    Darpit Dave, Steven R Steinhubl, R Brett McQueen.
      
    DOI:  https://doi.org/10.2337/dci25-0123
  10. Front Public Health. 2026 ;14 1735295
    Application study of an artificial intelligence and big data-based personalized chronic disease management model for diabetes patients.
    Mei Xin, Yanbing Yao, Ping Huang, Qiuxia Li.
       Background: To evaluate the real-world effectiveness of an artificial intelligence (AI) and big data-driven personalized chronic disease management model for type 2 diabetes mellitus (T2DM) patients, compared to conventional nurse-led management, and to identify factors associated with successful glycemic control within the personalized model.
    Methods: A retrospective cohort study was conducted involving 280 T2DM patients discharged from a single hospital between January 2019 and December 2024. Patients were divided into a conventional management group (n = 100) and a personalized management group (n = 180). The personalized group utilized a model integrating gradient boosting (XGBoost) for risk prediction and rule-based reasoning with reinforcement learning to dynamically generate individualized dietary, exercise, and blood glucose monitoring plans via a mobile application (APP). Both groups received 6 months of follow-up. Glycemic control [fasting blood glucose (FBG), 2-h postprandial glucose (2hPG), glycated hemoglobin (HbA1c)], self-care activities [Summary of Diabetes Self-Care Activities (SDSCA) scale], and quality of life [Diabetes-Specific Quality of Life (DSQL) scale] were assessed at baseline and 6 months. Within the personalized group, patients were further categorized into well-controlled (HbA1c ≤ 6.5%, n = 98) and poorly-controlled (HbA1c > 6.5%, n = 82) subgroups for case-control analysis.
    Results: At 6 months, the personalized management group demonstrated significantly better glycemic control (FBG: 6.79 ± 0.72 vs. 7.03 ± 0.89 mmol/L, p = 0.022; 2hPG: 6.27 ± 1.18 vs. 6.62 ± 1.16 mmol/L, p = 0.018; HbA1c: 6.48 ± 0.53% vs. 6.63 ± 0.46%, p = 0.018), superior self-care scores across all SDSCA domains (all p < 0.05, largest improvement in special diet: p = 0.001), and significantly higher quality of life (all DSQL dimensions p < 0.05) compared to the conventional group. Within the personalized group, multivariate analysis identified alcohol consumption [odds ratio (OR) = 3.576, p < 0.001], low baseline high-density lipoprotein cholesterol (HDL-C) (OR = 0.102, p = 0.007), and reduced blood glucose monitoring adherence (OR = 0.958, p < 0.001) as independent risk factors for poor control, while higher exercise plan completion was protective (OR = 0.976, p = 0.037).
    Conclusion: The AI and big data-driven personalized management model significantly improved glycemic control, self-care behaviors, and quality of life in T2DM patients over conventional care within 6 months. Success within the model is influenced by behavioral and biological factors, alongside alcohol consumption. This approach demonstrates promise for enhancing diabetes care.
    Keywords:  artificial intelligence; big data analytics; glycemic control; personalized diabetes management; quality of life; self-care activities
    DOI:  https://doi.org/10.3389/fpubh.2026.1735295
  11. JMIR Form Res. 2026 Feb 19. 10 e76936
    Smart Kiosk for Nutritional Management of People With Diabetes in Underserved Communities: Development and Technical Evaluation.
    Guadalupe Esmeralda Rivera-García, Juan Carlos Ramírez-Vázquez, Jaime Cruz-Casados, Miriam Janet Cervantes-López, Arturo LLanes-Castillo, Marco Antonio Diaz-Martinez.
       Background: Diabetes is a chronic disease with a high global prevalence, increasing from 200 million people in 1990 to 830 million in 2022, with a higher burden in low- and middle-income regions and high mortality in Mexico and Veracruz. These inequalities limit access to treatment and nutritional education, requiring technological solutions such as interactive kiosks based on artificial intelligence (AI) that contribute to the nutritional management of people with diabetes in marginalized communities.
    Objective: This study aimed to design and evaluate an interactive kiosk based on AI that generates culturally relevant and personalized meal plans for people with diabetes in marginalized communities.
    Methods: A low-cost prototype was developed, with a database of local foods and a multilayer perceptron trained with synthetic data based on national clinical guidelines. Performance was tested through an experimental evaluation that measured (1) the accuracy of nutritional recommendations compared with ideal meal plans (accuracy, precision, sensitivity, and F1-score); (2) performance, measured by recording response time with 1 to 50 simultaneous requests; and (3) usability, assessed using heuristic evaluation and the System Usability Scale (SUS).
    Results: The smart kiosk was experimentally evaluated in three dimensions: nutritional recommendations, system efficiency, and usability. The model achieved AI metrics of 87.3% overall accuracy, 90.5% precision, 92.1% sensitivity, and 91.3% F1-score. The average response time was 2.36 (SD 0.24) seconds in all load tests. A maximum time of 4 seconds was obtained in the simulation of 50 concurrent users. In the usability evaluation, an average score of 89 (SD 2.89) out of 100 was obtained on the SUS, which is considered excellent, along with a success rate of 98.3%.
    Conclusions: The AI-based kiosk demonstrated technical feasibility, adequate performance, and satisfactory usability. Its ability to operate without the need for internet and its low cost make it an equitable option for diabetes self-management and a replicable model in public health.
    Keywords:  artificial intelligence; diabetes management; nutritional recommendation system; smart health kiosk; underserved communities
    DOI:  https://doi.org/10.2196/76936
  12. Diabetes Metab Res Rev. 2026 Feb;42(2): e70139
    Algor-Ethics in Diabetes Care: Mapping the Route.
    Joshua Bemporad, Francesco De Domenico, Paolo Pozzilli.
      Diabetes mellitus represents a multifaceted global health challenge, frequently coexisting with obesity, cardiovascular complications, and metabolic disorders. Effective management requires individualised, evidence-based decisions informed by an array of clinical, genetic, and lifestyle data. With the rapid growth of digital health technologies, artificial intelligence (AI) and algorithmic systems have emerged as powerful tools to support clinicians in diagnosis, treatment planning, and risk stratification. While AI shows promise in improving diabetes outcomes and health system efficiency, its integration into patient care is not without ethical and epistemic challenges. Algorithmic decision-making can influence therapeutic strategies, sometimes without full transparency or adequate oversight, potentially compromising human values such as autonomy, justice, and trust. In this context, the discipline of 'Algor-ethics', a term coined to describe the intersection of algorithmic systems and ethical principles, becomes critical. This article explores the foundational concepts of Algor-ethics applied to diabetes care, analyzes the current state of AI integration, and highlights the epistemic and ethical implications of algorithmic decision-making. Emphasis is placed on developing a framework that ensures AI is implemented safely, equitably, and responsibly, particularly for complex patients with diabetes.
    Keywords:  Algor‐ethics; algorethics; artificial intelligence; diabetes; diabetes care; ethics; machine learning
    DOI:  https://doi.org/10.1002/dmrr.70139
  13. Analyst. 2026 Feb 16.
    Exploring the feasibility of near-infrared spectroscopy and machine learning for detecting cardiovascular diseases and diabetes mellitus in fingernails.
    Megan Wilson, Dhiya Al-Jumeily, Jason Birkett, Iftikhar Khan, Ismail Abbas, Matthew Harper, Sulaf Assi.
      Cardiovascular diseases (CVDs) and diabetes mellitus (DM) are significant conditions that impact lives around the globe. Frequently employed methods for detecting CVDs and/or DM such as blood work and cardiac catheterisation are often invasive, intrusive and can cause the patient additional physical and psychological harm. Vibrational spectroscopic methods including near-infrared (NIR) spectroscopy have emerged as novel methods for detecting medical conditions and diseases including amyotrophic lateral sclerosis, cancer, DM and periodontitis. NIR spectroscopy's ability to perform rapid and cost-effective analysis saves diagnostic waiting times, providing relief for strained healthcare systems. Moreover, their non-invasive, non-intrusive and non-destructive nature allow application to alternative biological matrices such as hair, fingernails and saliva. Therefore, this work explored the feasibility of NIR spectroscopy paired with machine learning (ML) for detecting CVDs and/or DM in fingernails. NIR spectroscopy successful characterised disease-related spectral features including key NIR regions related to the presence of advanced glycated end-products (AGEs), glycated proteins and DM. To further assess the detective capabilities of NIR spectroscopy, classification models were trained. Cubic and quadratic support vector machine (SVM) models demonstrated accuracy in terms of the classification of healthy, CVD and diabetic fingernails. Accuracy was further improved through binary classification models, which allowed the independent classification of CVD and DM spectra against healthy spectra. In summary, NIR spectroscopy combined with ML provided accurate detection for CVDs and DM in fingernails.
    DOI:  https://doi.org/10.1039/d5an01061f
  14. Front Public Health. 2026 ;14 1760871
    Evaluation of validity, reliability, and readability of AI chatbots for gestational diabetes mellitus: a multi-model comparative study.
    Xinxin Wang, Shuyan Lin, Hui Liu, Chuanqing Li, Li Zhou, Rongkang Li.
       Background: Gestational diabetes mellitus (GDM) is increasingly prevalent worldwide and is associated with substantial short- and long-term risks for mothers and offspring, making high-quality, accessible health information essential. At the same time, artificial intelligence (AI) chatbots based on large language models are being widely used for health queries, yet their accuracy, reliability and readability in the context of GDM remain unclear.
    Methods: We first evaluated six AI chatbots (ChatGPT-5, ChatGPT-4o, DeepSeek-V3.2, DeepSeek-R1, Gemini 2.5 Pro and Claude Sonnet 4.5) using 200 single-best-answer multiple-choice questions (MCQs) on GDM drawn from MedQA, MedMCQA and the Chinese National Medical Examination item bank, covering four domains: epidemiology and risk factors, clinical manifestations and diagnosis, maternal and neonatal outcomes, and management and treatment. Each item was posed three times to every model under a standardized prompting protocol, and accuracy was defined as the proportion of correctly answered questions. For public-facing information, we identified 15 core GDM education questions using Google Trends and expert review, and queried four chatbots (ChatGPT-5, DeepSeek-V3.2, Claude Sonnet 4.5 and Gemini 2.5 Pro). Two obstetricians independently assessed reliability using DISCERN, EQIP, GQS and JAMA benchmarks, and readability was quantified using ARI, CL, FKGL, FRES, GFI and SMOG indices.
    Results: Overall MCQ accuracy differed significantly across the six chatbots (p < 0.0001), with ChatGPT-5 achieving the highest mean accuracy (92.17%) and DeepSeek-V3.2 and Gemini 2.5 Pro performing comparably well, while ChatGPT-4o, DeepSeek-R1 and Claude Sonnet 4.5 scored lower. Newer model generations (ChatGPT-5 vs. ChatGPT-4o; DeepSeek-V3.2 vs. DeepSeek-R1) consistently outperformed their predecessors across all four domains. Among the four models evaluated on public-education questions, ChatGPT-5 achieved the highest reliability scores (DISCERN 42.53 ± 7.20; EQIP 71.67 ± 6.17), whereas Claude Sonnet 4.5, DeepSeek-V3.2 and Gemini 2.5 Pro scored lower. JAMA scores were uniformly low (0-0.07/4), reflecting poor transparency. All models produced text above the recommended sixth-grade reading level; ChatGPT-5 showed the most favorable readability profile (for example, FKGL 7.43 ± 2.42, FRES 62.47 ± 13.51) but still did not meet guideline targets.
    Conclusion: Contemporary AI chatbots can generate generally accurate and moderately reliable GDM-related information, with newer model generations showing clear gains in diagnostic validity. However, limited transparency and systematically high reading levels indicate that these tools are not yet suitable as stand-alone resources for GDM patient education and should be used as adjuncts to clinician counseling and professionally curated materials.
    Keywords:  artificial intelligence; gestational diabetes mellitus; large language models; patient education; readability
    DOI:  https://doi.org/10.3389/fpubh.2026.1760871
  15. Commun Med (Lond). 2026 Feb 17.
    A machine learning model for optimizing treatment of patients with poorly controlled type 2 diabetes.
    Juan Shi, Cong Liu, Jiahang Hu, Yuancheng Dai, Ying Peng, Fengmei Xu, Haonan Shi, Yunsong Li, Jun Li, Zunhai Zhou, Chunfang Wen, Shan Huang, Yi Shu, Xiaolin Ye, Aifang Wang, Hongxia Zhao, Ping Feng, Shengli Wu, Dandan Wang, Ping Liu, Yi Shi, Shuhui Yang, Weiwei Tu, Lei Chen, Chaoli Yu, Guang Ning, Yufan Wang, Zhiyun Zhao, Yifei Zhang, Weiqing Wang.
       BACKGROUND: Cardiorenal-protective sodium-glucose cotransporter-2 inhibitors (SGLT-2i) and glucagon-like peptide-1 receptor agonists (GLP-1RA) lack selection guidance. We aimed to build a SGLT-2i/GLP-1RA Decision Score (TiP DecScore) to tailor selection between them.
    METHODS: We developed the TiP DecScore in type 2 diabetes (T2D) patients receiving either therapy from the China Metabolic Analytics Project (derivation dataset: n = 24,322; validation dataset: n = 1,459), using gradient boosting decision tree and 15 features. The primary outcomes were glycated haemoglobin (HbA1c) control (<7%) and HbA1c levels at 6 and 12 months. The model's clinical effectiveness was evaluated by comparing HbA1c control between concordant (receiving the predicted optimal therapy) and discordant groups (receiving the predicted non-optimal therapy).
    RESULTS: Here we show the derivation cohort has mean (SD) age 53.7 (11.5) years, 63.0% males. Model validation shows good predictive performance (the receiver operating characteristic curve 0.71-0.78). GLP-1RA is favored over SGLT-2i (57.6% vs. 24.2% at 6 months; 57.9% vs. 28.6% at 12 months). At 6 months, compared with SGLT-2i, GLP-1RA is prioritized for patients with a shorter diabetes duration and higher fasting C-peptide, alanine aminotransferase, body mass index (BMI), and low-density lipoprotein cholesterol levels. At 12 months, patients with higher baseline HbA1c and BMI levels are more likely to be recommended GLP-1RA than SGLT-2i. Higher rates of HbA1c control are observed in concordant versus discordant groups, especially in younger patients (<55 years; 64.1% vs. 46.2%, P = 0.001) and males (58.6% vs. 45.6%, P = 0.018) at 12 months.
    CONCLUSIONS: The TiP DecScore effectively guides personalized selection between SGLT-2i and GLP-1RA therapies for T2D patients.
    DOI:  https://doi.org/10.1038/s43856-026-01442-8
  16. Ophthalmol Sci. 2026 Mar;6(3): 101064
    Re: Henriques et al Correspondence Re: Most et al: Can Multimodal Large Language Models Diagnose Diabetic Retinopathy from Fundus Photos? A Quantitative Evaluation.
    Jesse A Most, Dirk-Uwe G Bartsch, Shyamanga Borooah.
      
    DOI:  https://doi.org/10.1016/j.xops.2026.101064
  17. Healthc Technol Lett. 2026 Jan-Dec;13(1):13(1): e70060
    A Hybrid Ensemble Approach for Early-Stage Diabetes Detection.
    Rachana Katuwal Kc, Su Yang.
      Diabetes has become a critical global health concern, particularly in regions where access to diagnostic facilities is limited. In this work, we propose a hybrid framework that combines extreme gradient boosting (XGBoost) and deep neural networks (DNNs) for early-stage diabetes detection, using soft voting to generate the final ensemble predictions. The proposed framework was evaluated on two datasets: the widely used Diabetes UCI dataset and a newly collected dataset from Nepal. The ensemble method achieved 99% accuracy (ACC) with an area under the curve (AUC) of 1.00 on the Diabetes UCI dataset, and 91% ACC with a 0.96 AUC on the Nepal diabetes dataset, demonstrating its strong generalisability across distinct populations. Compared to individual models, the hybrid approach offered increased stability and a lower rate of false negatives, which is particularly important in clinical contexts. These findings highlight the potential of hybrid machine learning-deep learning models as cost-effective, scalable and generalisable decision-support tools for diabetes risk assessment.
    Keywords:  decision trees; diseases; health care; man‐machine systems
    DOI:  https://doi.org/10.1049/htl2.70060
  18. Ophthalmol Sci. 2026 Mar;6(3): 101063
    Re: Most et al: Can Multimodal Large Language Models Diagnose Diabetic Retinopathy from Fundus Photos? A Quantitative Evaluation.
    Lavinia Loss Henriques, Carolina Pelegrini Barbosa Gracitelli, Fernando Roberte Zanetti, Ricardo Luz Leitão Guerra.
      
    DOI:  https://doi.org/10.1016/j.xops.2026.101063
  19. Ophthalmol Sci. 2026 Jan;6(1): 100967
    Identifying Patients at High Risk of Lapses in Diabetic Retinopathy Care: A Machine Learning Study.
    Jizhou Tian, Diep Tran, Zainab Rustam, Gina Zhu, Paul Nagy, Hadi Kharrazi, Deidra C Crews, Zitong Wang, Scott L Zeger, Cindy X Cai.
       Objective: To predict lapses in diabetic retinopathy (DR) care.
    Design: Retrospective cohort study.
    Subjects: Adults ≥18 years with diabetes seen at the Wilmer Eye Institute for DR screening or treatment between 2012 and 2022.
    Main Outcome Measures: Whether an office visit for DR screening or treatment was followed by a lapse in care.
    Methods: Three versions of prediction algorithms were constructed using random forests (RFs). XGBoost (XGB) was used as a confirmatory analysis. Random forest-A and XGB-A included electronic health record (EHR) variables alone (e.g., sociodemographic, insurance, ophthalmic diagnoses, lead time, and recommended follow-up time). Random forest-B and XGB-B added location-based social determinants of health (SDoH) variables (e.g., Area Deprivation Index). Random forest-C and XGB-C added history of lapses in care (e.g., whether the patient has ever had lapses in care before). The area under the receiver operating characteristic curve (AUROC) and the area under the precision-recall curve (AUPRC) were calculated for each algorithm.
    Results: A total of 36 995 patients (mean age 62 years, 53% female, 47% non-Hispanic White, 38% non-Hispanic Black, and 4% Hispanic) and 141 930 office visits were included. The best performing model was RF-C with an AUROC of 0.774 (0.772-0.776) and AUPRC of 0.707 (0.704-0.711), outperforming RF-A and RF-B in AUROC and AUPRC (P < 0.001 for each comparison). XGB-C similarly outperformed XGB-A and XGB-B (P < 0.001 for each comparison).
    Conclusions: We developed RF algorithms, as well as XGB confirmatory models, to predict whether patients with diabetes will experience a lapse in DR care. The best prediction was achieved using EHR variables, location-based SDoH variables, and history of lapses in care. These models offer the opportunity to identify high-risk patients and offer additional resources to reduce lapses in care and potentially vision loss from DR.
    Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
    Keywords:  Diabetic retinopathy; Electronic health record; Lapses in care; Prediction; Social determinants of health
    DOI:  https://doi.org/10.1016/j.xops.2025.100967
  20. IEEE Trans Comput Biol Bioinform. 2026 Feb 16. PP
    Deep-Q-Network in Multiview Ensemble Learning to Predict Anti-diabetic Peptide and Diabetic Types.
    Aditya Kumar, Deepak Singh.
      Diabetes mellitus disease has become a significant human health concern globally, causing cardiovascular complications, kidney disease, cerebrovascular accidents, etc. In humans, bioactive molecules, specifically Antidiabetic peptides (ADPs), target $\beta$ or $T$-cells to regulate insulin production, making them potential alternatives for diabetes therapy. Recently, several computational ensemble methods have been investigated to predict ADPs. Notwithstanding progress in ensemble techniques, finding optimal solutions for predicting ADPs requires continuous effort for improvement. Therefore, the current study proposed a novel Deep-Q-Network (DQN)-based Reinforcement learning (RL) approach for optimizing a multiview ensemble to predict anti-diabetic peptides and diabetic types. Initially, to obtain the multiview, diverse feature encodings were employed, which were then systematically categorized into three heterogeneous groups based on the information sources and information-extracting techniques. Each group is considered as a view, and these views are based on sequence, physicochemical properties with sequence composition, and evolutionary information. Next, the proposed ensemble is constructed with 36 branch classifiers derived from 6 convolutional neural networks(CNNs). A reinforcement learning technique is employed to optimize the ensemble search space of $2^{36}-1$, made by thirty-six branch CNNs and three views. The DQN approach encodes 36-bit binary strings to find the optimal ensemble. The optimal ensemble with an ideal view-classifier pair enhances antidiabetic prediction efficacy. It surpasses existing state-of-the-art with a classification accuracy of 96.37%, a matthews correlation coefficient of 0.870, and an area under the curve of 0.983 on the standard ADPs dataset. Lastly, the performance of the proposed ensemble is considered to be its efficacy in peptide antidiabetic activity prediction.
    DOI:  https://doi.org/10.1109/TCBBIO.2026.3665266
This page is being maintained by Thomas Krichel. It was last updated on 2026‒02‒22 at 0:13.