bims-aukdir 2026-09-27 papers

bims-aukdir

Biomed News

on Automated knowledge discovery in diabetes research

Issue of 2026–09–27
sixteen papers selected by
Mott Given

Cross-fusion of digital twins and artificial intelligence in diabetes: from mechanistic elucidation to full-cycle precision management.
Machine learning-based risk prediction models for type 2 diabetes in primary care: a scoping review.
A Text-Guided Lesion Mining Vision-Language Ordinal Classification Framework for Diabetic Retinopathy Grading.
Performance of machine learning-based prediction models for hypoglycemia in Chinese patients with diabetes: A systematic review and meta-analysis.
Development and Validation of a Machine Learning-Based Model for Predicting All-Cause Mortality Risk in Patients With Type 2 Diabetes Mellitus Combined With Hypertension: National Cohort Study.
Artificial intelligence-driven diabetic retinopathy research: mapping the evolution, coupling, and global collaboration landscape (1996-2026).
A Hybrid and Comparative Machine Learning Framework for Predicting Gestational Diabetes Mellitus via a Prospective Case-Control Design: A Pilot Study Integrating Periodontal Health and Hematological Inflammatory Markers.
Interpretable Multihorizon Glucose Forecasting for Assessing Nutritional Information Impact in Type 1 Diabetes Management: Model Development and Validation Study.
Serum Iron as a Key Predictor of Diabetes Risk: Findings from 13 Cities in Saudi Arabia.
A multi-label deep learning system for simultaneous detection of nine fundus conditions.
Metabolite Signatures in Postpartum Prediabetes and Diabetes Among Indian Women With Prior Gestational Diabetes Mellitus: Insights From CHIP-F Study.
Artificial Intelligence-Driven Digital Tools for Diabetes Self-management in Children: A Scoping Review.
Metals and metalloids signatures in early pregnancy suggest cadmium and arsenic as major mixture components associated with gestational diabetes mellitus risk.
Stakeholder Perspectives on the Integration of AI in Diabetes Care: Systematic Review of Qualitative Studies.
Machine Learning-Assisted Terahertz Metasurface Sensing for Diabetes Progression.
A lipid metabolism-based gene signature defines diagnosis and molecular heterogeneity in diabetic nephropathy.

Front Endocrinol (Lausanne). 2026 ;17 1919404

Cross-fusion of digital twins and artificial intelligence in diabetes: from mechanistic elucidation to full-cycle precision management.

Yue Sun, Ruodi Yang, Qirui Xin, Xinxin Luo, Ying Zhou, Zishan Fu, Jiaxin Li, Zhichao Gu, Juntong Liu, Qingfeng Wang, Yufeng Yang.

  The integration of digital twins (DT) and artificial intelligence (AI) is driving a paradigm shift in diabetes mellitus (DM) care from traditional population-based, symptomatic intervention to a lifecycle-oriented precision management approach. This review systematically elaborates on the cross-disciplinary integration mechanisms of the two technologies. By integrating multimodal data and constructing physiologically constrained hybrid models, it achieves multi-scale mechanistic analyses ranging from single-cell β-cell protection to multi-organ complications at the basic research level. On the clinical application front, it not only significantly enhances the early screening sensitivity for diabetic retinopathy and the accuracy of blood glucose prediction, but also optimizes insulin dosing, personalized nutritional plans, and exercise decision support through virtual trials. In the field of drug development, virtual clinical trials accelerate target discovery and drug repurposing. Although substantial technical, regulatory, and ethical challenges remain unresolved, ongoing progress in hybrid modeling, federated learning, explainable AI, and evolving regulatory frameworks provides a plausible pathway toward individualized prediction and proactive management, provided that claims of clinical readiness are matched by rigorous prospective validation.

Keywords:  artificial intelligence; diabetes mellitus; digital twin; hybrid modeling; personalized management; precision medicine

DOI:  https://doi.org/10.3389/fendo.2026.1919404
Front Public Health. 2026 ;14 1946070

Machine learning-based risk prediction models for type 2 diabetes in primary care: a scoping review.

Rashita Ravi, Jeby Jose Olickal.

   Background: Type 2 diabetes mellitus (T2DM) is a major global public health challenge, with many individuals remaining undiagnosed until complications develop. Machine learning (ML)-based risk prediction models have the potential to support early identification of individuals at increased risk using primary care data. However, the characteristics and applicability of these models within primary care settings have not been comprehensively mapped.
Objective: To systematically map the available evidence on machine learning (ML)-based models for risk prediction, early detection, and case-finding of type 2 diabetes in primary care, and to summarize their characteristics, including predictors, modeling approaches, validation strategies, and model performance.
Methods: A scoping review was conducted following the Joanna Briggs Institute methodology and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR). PubMed/MEDLINE, Scopus, and Ovid MEDLINE were searched for English-language studies published between January 2011 and December 2025. Primary research describing ML-based risk prediction models developed, validated, evaluated, or intended for implementation in primary care was eligible. Data were extracted using a structured charting form and synthesized descriptively.
Results: The search identified 186 records, of which four studies met the inclusion criteria. The included studies were conducted in Sweden, Canada, Saudi Arabia, and Hong Kong between 2024 and 2025. Three studies focused on model development and internal validation, while one externally validated previously developed non-laboratory prediction models. A range of ML approaches was identified, including stochastic gradient boosting, federated learning, multilayer perceptron, random forest, support vector classification, naïve Bayes, and decision tree algorithms, with logistic regression commonly used as a comparator. Models primarily utilized routinely collected demographic, anthropometric, lifestyle, and electronic health record-derived variables. Most studies reported moderate-to-good predictive performance; however, evidence regarding external validation, calibration, and prospective implementation within routine primary care remained limited.
Conclusion: Evidence on ML-based risk prediction models for T2DM applicable to primary care remains limited despite growing interest in AI for diabetes prediction. Existing models demonstrate promising predictive performance using routinely available clinical information, but greater emphasis is needed on external validation, calibration, prospective implementation, and evaluation across diverse primary care populations before widespread clinical adoption.
Review registration: Open Science Framework https://osf.io/mbfrz.

Keywords:  artificial intelligence; machine learning; primary care; risk prediction; type 2 diabetes mellitus

DOI:  https://doi.org/10.3389/fpubh.2026.1946070
J Imaging. 2026 Sep 01. pii: 412. [Epub ahead of print]12(9):

A Text-Guided Lesion Mining Vision-Language Ordinal Classification Framework for Diabetic Retinopathy Grading.

Jiawen Huang, Jinxia Shang.

  Diabetic retinopathy (DR) is a major retinal disease that can cause visual impairment and irreversible blindness. Accurate automated DR grading is essential for large-scale screening and timely clinical intervention. However, most existing methods rely primarily on visual features for classification. Moreover, they often overlook the ordinal structure of DR severity and the intra-class phenotypic heterogeneity arising from diverse lesion combinations. To address these issues, based on the semantic prior information provided by RetiZero, we propose a text-guided lesion mining vision-language ordinal classification framework for DR grading. The proposed framework introduces a text-guided cross-layer lesion mining module that exploits semantic response differences between normal-tissue and lesion-related textual prompts, thereby guiding multi-level visual patch features toward lesion regions relevant to DR grading. To explicitly model the ordered progression of DR severity, we design a conditional ordinal regression branch and an ordinal distribution alignment strategy that jointly encourage the predictions to follow the inherent order of DR grades. Moreover, we introduce a multi-center feature constraint to capture diverse intra-grade phenotypic patterns and enhance feature discriminability. Experiments on APTOS 2019 show that the proposed method achieves 86.3% accuracy, 90.6% AUC, and 70.9% Macro-F1, which improved by 2.4, 0.7, and 5.5 percentage points compared to RetiZero. Furthermore, under the standardized leave-one-domain-out protocol of GDRNet, the proposed method achieves the highest reported average accuracy of 58.6% across six public DR datasets, exceeding the reported result of PAF (54.6%) by 4.0 percentage points. Nevertheless, our approach is limited in F1 and AUC metrics, for which GDRNet delivers superior performance. These results suggest that the proposed framework can improve DR grading performance and the cross-dataset generalization ability of the model to a certain extent.

Keywords:  cross-layer lesion mining; diabetic retinopathy; multi-center feature constraint; ordinal distribution alignment; vision–language model

DOI:  https://doi.org/10.3390/jimaging12090412
PLoS One. 2026 ;21(9): e0358876

Performance of machine learning-based prediction models for hypoglycemia in Chinese patients with diabetes: A systematic review and meta-analysis.

Jinhua Yan, Yanping Song, Yangmei Du, Fanmin Li.

OBJECTIVES: This study aimed to systematically evaluate the predictive performance of machine learning (ML)-based models for predicting hypoglycemia in Chinese patients with diabetes.
METHODS: We systematically searched PubMed, Embase, Web of Science, the Cochrane Library, CINAHL, CNKI, and Wanfang databases from inception to February 2026. Eligible studies focused on the development or validation of ML-based models for predicting hypoglycemia in Chinese patients with diabetes. Study selection and data extraction were performed independently by two reviewers. Information on study characteristics, modeling approaches, predictors, validation methods, and model performance was collected. The area under the receiver operating characteristic curve (AUC) was synthesized using a random-effects model. Study quality was assessed using the Prediction Model Risk of Bias Assessment Tool (PROBAST).
RESULTS: A total of 13 studies were included, and the pooled prevalence of hypoglycemia was 25% (95% CI: 17%-33%). The overall pooled area under the receiver operating characteristic curve (AUC) was 0.90 (95% CI: 0.87-0.93). Subgroup analyses by modeling algorithms showed pooled AUCs of 0.89 for extreme gradient boosting (XGBoost), 0.88 for random forest (RF), 0.85 for support vector machine (SVM), 0.84 for Light Gradient Boosting Machine (LightGBM), 0.83 for logistic regression (LR), and 0.81 for decision tree (DT) models. Common predictors included age, insulin use, body mass index, HbA1c, creatinine, and history of hypoglycemia.
CONCLUSION: We attempted to provide a comprehensive overview of machine learning-based prediction models for hypoglycemia in patients with diabetes. Research in this field remains at an early stage, although several models with good discriminatory performance have been reported. Methodological limitations and insufficient validation were observed in many studies. Concerns regarding model robustness and interpretability also exist. More efforts to develop reliable and interpretable models and to promote their application in clinical practice for early risk identification are needed.

DOI: https://doi.org/10.1371/journal.pone.0358876
JMIR Med Inform. 2026 Sep 22. 14 e85557

Development and Validation of a Machine Learning-Based Model for Predicting All-Cause Mortality Risk in Patients With Type 2 Diabetes Mellitus Combined With Hypertension: National Cohort Study.

Wenlong Ding, Lei Fang, Cunming Fang.

   BACKGROUND: Type 2 diabetes mellitus (T2DM) combined with hypertension significantly increases mortality risk, yet accurate risk prediction models remain limited.
OBJECTIVE: We aimed to develop and validate machine learning-based models to predict all-cause mortality in patients with T2DM and hypertension.
METHODS: We analyzed data from the National Health and Nutrition Examination Survey from 1999 to 2018 linked with mortality data up to December 31, 2019. Adult participants (aged ≥20 years) with concurrent T2DM and hypertension were included. Five machine learning algorithms were developed and compared: random forest, light gradient boosting machine, decision tree, extreme gradient boosting, and logistic regression. Model performance was evaluated using area under the curve (AUC), calibration plots, and decision curve analysis.
RESULTS: A total of 2428 participants were included (mean age 62.05, SE 0.33 years; n=1218, 50.15% female). During a median follow-up of 6.75 (IQR 4.20-8.90) years, among the 2428 patients, 719 (29.6%) deaths occurred. The random forest model demonstrated superior performance (AUC=0.873, 95% CI 0.856-0.891) compared to light gradient boosting machine (AUC=0.785), decision tree (AUC=0.732), extreme gradient boosting (AUC=0.792), and logistic regression (AUC=0.783). Key predictive features included age, race, chronic kidney disease, BMI, and blood urea nitrogen. The model exhibited excellent calibration and clinical utility across various risk thresholds.
CONCLUSIONS: Our machine learning-based model provides accurate all-cause mortality prediction for patients with T2DM and hypertension, potentially supporting clinical decision-making and risk stratification in this high-risk population.

Keywords:  AI; NHANES; National Health and Nutrition Examination Survey; artificial intelligence; hypertension; machine learning; mortality prediction; risk stratification; type 2 diabetes mellitus

DOI:  https://doi.org/10.2196/85557
Front Endocrinol (Lausanne). 2026 ;17 1935040

Artificial intelligence-driven diabetic retinopathy research: mapping the evolution, coupling, and global collaboration landscape (1996-2026).

Yihui He, Danyu Li, Danbing Li, Yunci Ma, Wentao Huang.

   Background: Diabetic Retinopathy (DR) is a major global cause of blindness. Artificial Intelligence (AI) has markedly impacted fundus screening over the past three decades, yet systematic bibliometric mapping of the knowledge architecture, evolutionary paths, and synergy among AI models, data modalities, and DR remains scarce.
Objective: This study conducted a comprehensive bibliometric analysis to characterize the AI-driven DR knowledge structure, collaboration networks, and hotspot migration, and to elucidate co-evolutionary dynamics among AI advances, data modality development, and DR research.
Methods: Following a systematic search and screening process, 12,741 publications were identified from the Web of Science Core Collection (WoSCC), PubMed, and Scopus, spanning from January 1996 to June 2026. The analytical framework combined multiple methodologies: VOSviewer for collaborative network visualization, CiteSpace for burst detection and timeline mapping, and a Python-based text-mining pipeline for standardized extraction and normalization of AI model names, data modalities, and disease entities.
Results: The field exhibits a distinct three-stage evolutionary trajectory: the traditional machine learning era (1996-2014), the deep learning surge (2015-2019), and the current phase marked by the growing prominence of Transformer-based models (2020-present). The collaboration landscape is multipolar, with the United States, China, and India as hubs, while Singapore produces high-impact research. The knowledge base rests on algorithmic innovation and clinical validation. Convolutional neural networks have long served as the backbone architecture in the literature, while Vision Transformers have shown a clear upward trend in publication volume in recent years. Research hotspots are expanding from single-disease classification toward multimodal integration. Although fundus imaging remains the predominant data source, the potential of electronic health record narratives and multi-omics data is increasingly recognized. Overall, the research focus is shifting from "black-box" pattern recognition toward explainable AI and end-to-end clinical translation.
Conclusion: This study presents a systematic bibliometric mapping of AI-driven DR research, revealing high-frequency co-occurrence patterns between architectural specialization and clinical demands. Challenges persist in data integration, rare-disease evidence, and cross-setting validation. The future is likely to be shaped by multimodal foundation models and portable acquisition, transitioning AI toward comprehensive clinical decision support.

Keywords:  artificial intelligence; bibliometrics; deep learning; diabetic retinopathy; knowledge graph; machine learning; multimodal fusion

DOI:  https://doi.org/10.3389/fendo.2026.1935040
Metabolites. 2026 Aug 27. pii: 617. [Epub ahead of print]16(9):

A Hybrid and Comparative Machine Learning Framework for Predicting Gestational Diabetes Mellitus via a Prospective Case-Control Design: A Pilot Study Integrating Periodontal Health and Hematological Inflammatory Markers.

İsa Temur, Mehmet Özsan, Katibe Tuğçe Temur, Andaç Batur Çolak.

  Background/Objectives: This pilot study aimed to develop and compare advanced machine learning models, specifically a Bayesian-regularized artificial neural network (ANN) and a transformer-enhanced physics-informed neural network (PINN), by integrating periodontal health indices and hematological inflammatory markers for the prediction of GDM. Methods: Utilizing a prospective case-control study design, a clinical dataset comprising 80 pregnant women (40 with GDM and 40 healthy controls) was evaluated to develop and compare advanced machine learning models. Clinical, periodontal, and complete blood count-derived inflammatory parameters were integrated into two predictive models: a Bayesian regularization-based artificial neural network (ANN) and a transformer-enhanced physics-informed neural network (PINN). Model performance was evaluated using the coefficient of determination (R2), mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), and residual error analyses. Results: The ANN demonstrated superior predictive performance, achieving an R2 of 0.9872 and an MSE of 3.38 × 10-3, compared with an R2 of 0.9833 for the PINN model. Residual analysis showed that the ANN provided greater prediction stability, with a mean deviation of 0.5691 and a standard deviation of 3.6400. The findings demonstrate that the integrated analysis of periodontal and hematological markers through advanced neural architectures, particularly the Bayesian-regularized ANN, provides a high-fidelity diagnostic signal for GDM prediction. This integrated feature set, when processed by the proposed machine learning framework, enables the identification of complex biological patterns with remarkable precision. Conclusions: The proposed machine learning framework provides an accurate, non-invasive, and clinically applicable approach for predicting GDM. Integrating routinely available periodontal and hematological data may improve early risk stratification and support personalized prenatal care. Further validation in larger and more diverse populations is warranted before clinical implementation.

Keywords:  artificial neural network; complete blood count; gestational diabetes mellitus; inflammatory biomarkers; machine learning; periodontal disease; precision medicine; pregnancy; risk prediction

DOI:  https://doi.org/10.3390/metabo16090617
JMIR AI. 2026 Sep 21. 5 e105049

Interpretable Multihorizon Glucose Forecasting for Assessing Nutritional Information Impact in Type 1 Diabetes Management: Model Development and Validation Study.

Carlos Gallardo-García, M Elena Hernando, David Subías, Mercedes Rigla, Gema García-Sáez.

   Background: Type 1 diabetes is characterized by absolute insulin deficiency, requiring exogenous insulin therapy to maintain blood glucose levels within safe ranges. Postprandial glucose control remains particularly challenging, and current meal-related strategies are mainly based on carbohydrate intake. However, other macronutrients, such as fats and proteins, may also influence the magnitude and timing of the glycemic response and are not usually incorporated into glucose forecasting models.
Objective: This study aims to develop and evaluate multihorizon blood glucose forecasting models to assess the impact of incorporating detailed nutritional information, with particular emphasis on the postprandial period and to analyze how the contribution of different nutrients varies across prediction horizons.
Methods: We trained temporal fusion transformer (TFT) models on continuous glucose monitoring, insulin, and meal data with different combinations of nutritional variables from 351 adults with type 1 diabetes (15,601 meals) to predict glucose values up to 4 hours ahead. Carbohydrates were used as the baseline nutritional input, and additional configurations included fats, proteins, sugars, and complex carbohydrates. Model performance was evaluated globally and in predictions initiated at meal intake. Prediction horizons were grouped into early and late intervals, corresponding to 0-2 hours and 2-4 hours, respectively. The interpretability mechanisms of the TFT and integrated gradients method were used to analyze the relative contribution of nutritional variables.
Results: Models incorporating additional nutritional information generally outperformed the carbohydrate-only baseline. In the global analysis, the combination of carbohydrates, fats, and proteins achieved the best performance in the early prediction interval, reducing mean root mean squared error (RMSE) and mean absolute error (MAE) from 30.48 mg/dL and 21.02 mg/dL to 29.23 mg/dL and 20.15 mg/dL. At late intervals, distinguishing between complex carbohydrates and sugars, together with fats and proteins, provided the best performance, reducing mean RMSE and MAE from 45.34 mg/dL and 34.73 mg/dL to 43.24 mg/dL and 32.65 mg/dL. A similar temporal dependency pattern was observed in the postprandial evaluation and was partially supported by the interpretability analysis.
Conclusions: These findings suggest that incorporating a more comprehensive representation of meal composition may modestly improve postprandial glucose forecasting. Horizon-specific attribution patterns indicated that the models used nutritional inputs differently across the forecast horizon; however, these findings should not be interpreted as causal or physiological effects of individual nutrients. Further external and prospective validation is required before the potential clinical utility of these models can be assessed.

Keywords:  attention model; clinical decision-support; deep learning; explainable artificial intelligence; glucose forecasting; multihorizon prediction; nutritional information; temporal fusion transformer; type 1 diabetes

DOI:  https://doi.org/10.2196/105049
Metab Syndr Relat Disord. 2026 Sep 21. 15578518261479845

Serum Iron as a Key Predictor of Diabetes Risk: Findings from 13 Cities in Saudi Arabia.

Saeed Awad M Alqahtani, Fadwa M Alkhulaifi, Jamilah Alshammari, Rasha Alonaizan, Sheka Yagub Aloyouni, Nada Bawazir, Hadiah B Almahdi, Zuhier Awan, Suliman Alomar.

  Diabetes mellitus represents a major public health challenge in Saudi Arabia. This study investigated the associations of gender, age, body mass index (BMI), and serum iron levels with diabetes risk across 13 Saudi cities and developed a machine-learning model for diabetes prediction. Data from 49,259 individuals were obtained from Al Borg Laboratories for the period 2015-2023. Descriptive statistics, independent-samples t-tests, chi-squared tests, and correlation analyses were performed. Logistic Regression, Random Forest, and Gradient Boosting models were evaluated using gender, age, BMI, and serum iron as predictors. The best-performing model was optimized through grid search with five-fold cross-validation. The cohort had a mean age of 45.92 years, mean BMI of 27.33 kg/m2, and mean serum iron concentration of 87.11 µg/dL. The prevalence of diabetes was 10.31% and was higher among males than females (11.42% vs. 9.24%). Individuals with diabetes had significantly lower serum iron concentrations than those without diabetes (79.76 vs. 88.26 µg/dL; P < 0.001), and gender was significantly associated with diabetes status (P < 0.001). Random Forest demonstrated the best predictive performance, with an accuracy of 0.92, precision of 0.90, recall of 0.94, and area under the receiver operating characteristic curve of 0.96. Following hyperparameter tuning, the optimized model achieved a test-set accuracy of 0.90. Overall, gender, age, BMI, and serum iron levels were significantly associated with diabetes in the Saudi population, with lower serum iron levels linked to a higher prevalence of the disease. The optimized Random Forest model may provide a useful tool for diabetes risk prediction, supporting the potential application of machine learning in early detection and prevention. These findings also highlight the importance of monitoring iron status and developing tailored strategies for individuals with diabetes.

Keywords:  Saudi Arabia; body mass index; diabetes mellitus; machine learning; predictive risk; serum iron

DOI:  https://doi.org/10.1177/15578518261479845
Asia Pac J Ophthalmol (Phila). 2026 Sep 25. pii: S2162-0989(26)00105-2. [Epub ahead of print] 100379

A multi-label deep learning system for simultaneous detection of nine fundus conditions.

Jungwoo Ha, Yeon Hee Choi, Danbi Lee, Junseo Choi, Wonyoung Seo, Geunyoung Lee, Dongjin Nam, Sahil Thakur, Tae Keun Yoo, Sunjin Hwang.

   PURPOSE: To develop and validate a deep learning model for simultaneous detection of nine fundus conditions from a single color fundus photograph.
METHODS: A development dataset of 236,242 color fundus images was assembled from 17 heterogeneous sources and partitioned at the patient level into training (70%), tuning (10%), and internal validation (20%) sets; 5-fold cross-validation was applied on the training set. Nine target conditions were labeled using standardized or photographic criteria: diabetic retinopathy (DR), age-related macular degeneration (AMD), myopic macular degeneration (MMD), glaucoma suspect (GS), epiretinal membrane (ERM), retinal vascular occlusion (VO), media opacity, retinal hemorrhages, and any retinal disorder (composite). A multitask ConvNeXt architecture was trained end-to-end, with operating thresholds pre-specified via Youden's index on the tuning set. External validation was performed on an independent cohort of 4,055 images from an East Asian health-screening center, excluded from model development.
RESULTS: On internal validation (N = 47,229), AUROCs ranged from 0.943 (AMD) to 0.989 (VO), with sensitivity of 88.1%-97.1% and NPV ≥98.8% for eight of nine conditions. On independent external validation (N = 4,055), AUROCs ranged from 0.894 (retinal disorder) to 0.983 (MMD), with NPV ≥95.6% for eight of nine conditions, including 99.3% for DR, 99.9% for VO, and 99.8% for MMD.
CONCLUSIONS: This study presents a validated, criteria-anchored multi-label fundus triage framework capable of simultaneously screening for nine conditions from a single fundus photograph, with performance maintained on external validation at a source-excluded health-screening site within the same national health-screening system.

Keywords:  Artificial Intelligence; Deep learning; Diabetic retinopathy; Glaucoma; Myopic macular degeneration

DOI:  https://doi.org/10.1016/j.apjo.2026.100379
Diabetes Metab Res Rev. 2026 Sep;42(6): e70232

Metabolite Signatures in Postpartum Prediabetes and Diabetes Among Indian Women With Prior Gestational Diabetes Mellitus: Insights From CHIP-F Study.

Sahil Kapoor, Ambika Singh, Neelam Mahlawat, J Manivannan, Arpita Singh, Shinjini Bhatnagar, Nikhil Tandon, Yashdeep Gupta, Pallavi Kshetrapal.

   AIMS: Gestational diabetes mellitus (GDM) increases the risk of postpartum dysglycaemia. This study investigates candidate metabolites and perturbed metabolic pathways associated with postpartum prediabetes/diabetes among high-risk Indian/South-Asian women with prior GDM, a largely unexplored population.
MATERIALS AND METHODS: In this nested case-control study within the Cohort of Indian Women with Hyperglycaemia in Pregnancy and their Families (CHIP-F), untargeted metabolomics (UPLC-MS/MS) was performed on plasma samples from 216 women with prior GDM, which included 22 participants with postpartum diabetes, 88 with prediabetes and 106 with normoglycaemia.
RESULTS: Untargeted metabolomics combined with machine learning approach revealed distinct metabolic alterations associated with postpartum dysglycaemia (prediabetes/diabetes) among women with prior GDM. After adjustment for age, body mass index, family history of diabetes, and postpartum duration, each 1-standard deviation increase in the log-transformed concentrations of Tyrosine (aOR = 1.71, 95% CI: 1.01-2.91) and L-Kynurenine (aOR = 1.69, 1.02-2.80) was associated with higher odds of diabetes. Similarly, higher concentrations of 2-Aminoisobutyric acid (aOR = 1.50, 1.09-2.06), Butyrylcarnitine (aOR = 1.99, 1.44-2.75), L-Norleucine (aOR = 1.96, 1.41-2.72), and Propionylcarnitine (aOR = 1.94, 1.41-2.67) were associated with higher odds of prediabetes. Pathway analysis revealed potential dysregulation of aromatic amino acids (AAAs) metabolism in prediabetes. In diabetes, in addition to dysregulation of AAAs, potential lipid metabolic disturbances were also present.
CONCLUSIONS: This study identified distinct candidate metabolic signatures associated with postpartum prediabetes and diabetes in Indian/South-Asian women with a history of prior GDM, highlighting early disruptions in AAAs and lipid metabolism.

Keywords:  GDM; T2DM; biomarker; machine learning; metabolomics; pathophysiology; postpartum

DOI:  https://doi.org/10.1002/dmrr.70232
Int J Endocrinol Metab. 2027 Jan 31. 25(1): e169084

Artificial Intelligence-Driven Digital Tools for Diabetes Self-management in Children: A Scoping Review.

Maryam Nakhaee Moghadam, Sara Amini, Mojtaba Lotfi, Mehrnaz Nazari Rad, Jalal Nourmohammadi.

   Context: Managing diabetes in children is challenging and requires continuous monitoring and structured self-care support. Artificial intelligence (AI)-driven digital tools may enhance self-management and improve health outcomes. This scoping review aimed to synthesize the existing evidence on AI-driven digital tools for pediatric diabetes self-management, focusing on their functionalities, benefits, and limitations.
Evidence Acquisition: From an initial pool of 97 articles published between January 2010 and March 1, 2025, 19 studies met the predefined inclusion criteria. A systematic search was conducted in PubMed, Scopus, Web of Science, CINAHL, PsycINFO, and Embase using predefined keywords. Two independent reviewers performed screening and data extraction. Methodological quality was appraised using the Cochrane Risk of Bias Tool, the Newcastle-Ottawa Scale, and the Critical Appraisal Skills Programme (CASP) Checklist. Qualitative and quantitative findings were synthesized thematically.
Results: The 19 included studies addressed several overlapping application domains of AI-driven digital tools for pediatric diabetes self-management. Overall, 8 studies focused primarily on glucose prediction, glycemic monitoring, or pattern recognition; 4 studies evaluated insulin dosing optimization and clinical decision-support systems; 6 studies investigated personalized education, behavioral support, user engagement, or self-management learning interventions; 5 studies examined telemedicine, remote monitoring, or healthcare access; and 5 studies primarily addressed ethical, methodological, transparency, or implementation challenges.
Discussion: AI-based digital tools have considerable potential to support diabetes self-management in children. Future research should prioritize long-term evaluations, the inclusion of diverse populations, and rigorous assessments of clinical outcomes, data privacy, and algorithm performance.

Keywords:  Artificial Intelligence; Child; Diabetes Mellitus; Digital Health; Self-management

DOI:  https://doi.org/10.5812/ijem-169084
Environ Pollut. 2026 Sep 25. pii: S0269-7491(26)01608-8. [Epub ahead of print] 129238

Metals and metalloids signatures in early pregnancy suggest cadmium and arsenic as major mixture components associated with gestational diabetes mellitus risk.

Jingxia Wei, Zhanyue Zheng, Yingjie Zhou, Shuchang Liu, Yu Feng, Tianao Sun, Yongjie Ma, Minglian Pan, Xinyu Yuan, Jinhao Wan, Yan Sun.

  Pregnancy is a vulnerable window for environmental pollutant exposure, yet the contribution of metals and metalloids (metal(loid)s) to gestational diabetes mellitus (GDM) remains insufficiently characterized. This prospective cohort study evaluated whether early-pregnancy urinary metal(loid)s were associated with subsequent GDM and explored their predictive and mechanistic value. Among 1,201 pregnant women included in Guilin, China, nine urinary metal(loid)s were measured by inductively coupled plasma mass spectrometry. GDM was diagnosed by a 75-g oral glucose tolerance test at 24-28 gestational weeks. Restricted cubic spline and weighted quantile sum regression assessed dose-response and mixture effects. Machine learning models evaluated early prediction, and network toxicology explored molecular pathways. GDM occurred in 312 participants. Cadmium, arsenic, manganese, thallium, and lead showed nonlinear associations with GDM risk. Cadmium and arsenic dominated the overall mixture effect. Adding urinary metal(loid)s to conventional biochemical indicators modestly improved discrimination, with the XGBoost performing best. Network toxicology identified shared cadmium/arsenic targets enriched in oxidative stress, inflammation, apoptosis, endothelial dysfunction, and insulin-resistance pathways, including AGE-RAGE, HIF-1, TNF, and IL-17 signaling. These findings suggest that early-pregnancy metal(loid)s mixture exposure, particularly cadmium and arsenic, is associated with GDM risk, with arsenic the most consistent contributor across mixture methods and non-linear signals for lead and manganese warranting further investigation. Urinary metal(loid)s may serve as complementary biomarkers for early GDM risk stratification.

Keywords:  Gestational diabetes mellitus; Machine learning; Metals and metalloids; Network toxicology

DOI:  https://doi.org/10.1016/j.envpol.2026.129238
J Med Internet Res. 2026 Sep 22. 28 e105329

Stakeholder Perspectives on the Integration of AI in Diabetes Care: Systematic Review of Qualitative Studies.

Héctor Martínez-Martínez, Julia Martínez-Alfonso, Belén Sánchez-Rojo-Huertas, María Eugenia Visier-Alfonso, Fernando Sebastián-Valles, Ana Díez-Fernández, Ana Pérez-Moreno, Vicente Martínez-Vizcaíno.

   Background: AI is increasingly being integrated into diabetes care, with growing evidence supporting its potential to improve clinical decision-making, risk prediction, and self-management. However, the lived experiences, expectations, and concerns of those involved in its implementation have not been adequately synthesized.
Objective: This study aims to synthesize qualitative evidence on the perspectives of patients, caregivers, health care professionals (HCPs), and other stakeholders regarding the integration of AI in diabetes management.
Methods: We searched MEDLINE via PubMed, Web of Science, Scopus, CINAHL, and PsycINFO from inception to February 17, 2026. Eligible studies examined the perspectives of adult patients, caregivers, health care professionals, or administrators on AI-enabled tools for diabetes management, self-management, clinical decision support, or the prevention of complications, and used qualitative methods or reported a separately analyzable qualitative component. Tools required an identifiable data-driven function for prediction, classification, recommendation, personalization, or decision support. Studies focused exclusively on image-based diagnosis, technical validation, or digital tools without an identifiable AI component were excluded. Two reviewers (HM-M and JM-A) independently screened studies and extracted data. Methodological limitations were assessed using the Joanna Briggs Institute (JBI) checklist and the Cochrane Qualitative Methodological Limitations Tool (CAMELOT). Findings were synthesized using thematic synthesis. Confidence was assessed using GRADE-CERQual. A sensitivity analysis excluded questionnaire-based qualitative evidence.
Results: Fourteen studies published between 2023 and 2025 were included, representing at least 738 participants across 9 countries. Participants included individuals with type 1 or type 2 diabetes, family caregivers, doctors, nurses, specialists, administrators, and other health care staff. Studies evaluated large language models, AI-enabled mobile applications and wearables, glucose-prediction systems, and clinical decision-support tools. Exposure ranged from direct use of functioning systems to evaluation of prototypes, wireframes, and hypothetical applications. Four analytical themes and 13 subthemes were identified. Stakeholders perceived that AI could support preventive and individualized care, education, self-management, and decision-making. Concerns included accuracy, bias, privacy, accountability, increased workload, caregiver burden, loss of professional autonomy, and erosion of human-centered care. Participants emphasized explainability, intuitive design, integration with existing systems, tailored training, and continued access to human support. Eleven findings were rated as high confidence and 2 as moderate confidence.
Conclusions: Although stakeholders perceived AI to be useful for diabetes care, these qualitative findings do not demonstrate clinical effectiveness, safety, or improved patient outcomes. Evidence was limited by heterogeneity in AI modalities, stakeholder groups, diabetes contexts, and technology exposure, as well as demographic imbalance, restricted reporting of researcher reflexivity, and reliance on prototype or hypothetical systems. Implementation should prioritize transparent design, clinical validation, data governance, human oversight, and tailored support, while preserving professional judgment and person-centered relationships.

Keywords:  artificial intelligence; diabetes mellitus; digital health; qualitative research; stakeholder perspectives; trust

DOI:  https://doi.org/10.2196/105329
IEEE Trans Nanobioscience. 2026 Sep 14. PP

Machine Learning-Assisted Terahertz Metasurface Sensing for Diabetes Progression.

Niten Kumar Panda, Sananda Kumar.

A polarization-insensitive terahertz (THz) metasurface refractive index sensor is proposed for plasma amino acid-based detection of type 2 diabetes progression stages. The unloaded sensor exhibits a resonance frequency of 0.6638 THz at maximum absorptivity, with a high quality factor(Q) of 41.4, a figure of merit(FoM) of 4.17 and maximum refractive index sensitivity(S) of 66.67 GHz/RIU, enabling precise refractive index sensing. The combined effect of six clinically accepted plasma amino acid biomarkers is considered to model diabetes progression. Five progressive stages-Normal (N), Intermediate-1 (I-1), Intermediate-2 (I-2), Diabetic (D), and High-Diabetic (HD)-are characterized by variations in amino acid concentrations that alter the effective refractive index of blood plasma. The Lorenz-Lorentz formulation is employed to estimate stage-dependent refractive indices from molecular composition and density variations. The corresponding resonance shifts and absorptivity responses are obtained through parametric simulations and used to train multiple machine learning models. The proposed framework Ensemble Bagged Trees classifier achieves a classification accuracy of 94% in accurately identifying the precise stage of diabetes on the simulated test datasets. The results demonstrate the potential of integrating THz metasurface sensing with data-driven modeling for minimally invasive, stage-specific diabetes diagnostics.

DOI: https://doi.org/10.1109/TNB.2026.3733255
Iran J Basic Med Sci. 2026 ;29(10): 1510-1520

A lipid metabolism-based gene signature defines diagnosis and molecular heterogeneity in diabetic nephropathy.

Qingbu Mei, Xinyu Liu, Jing Xu, Shuang Li, Tao Wang, Hui Yuan.

   Objectives: To identify lipid metabolism-related biomarkers of diabetic nephropathy (DN), develop and validate a diagnostic model, and further explore immune infiltration patterns and molecular subtypes of DN.
Materials and Methods: Multiple public transcriptomic datasets were integrated to identify lipid metabolism-related biomarkers for DN using differential expression analysis, weighted gene co-expression network analysis, and machine learning. A diagnostic model was then constructed from these biomarkers, and its performance was evaluated using ROC curve and nomogram analyses, followed by validation in an independent cohort. In addition, immune cell infiltration was assessed with the CIBERSORT algorithm, and consensus clustering identified molecular subtypes of DN. Finally, the findings were validated in a model of type 2 diabetes.
Results: A total of 28 lipid metabolism-related differentially expressed genes were identified, and five hub biomarkers (G0S2, PTGDS, CA2, CYP27B1, and HSD17B14) were screened and demonstrated favorable diagnostic performance in both training and validation cohorts. Functional enrichment analysis revealed that these genes were mainly involved in fatty acid metabolism and PPAR signaling pathways. Immune infiltration analysis revealed significant correlations between dysregulated lipid metabolism and immune cell infiltration in DN. Consensus clustering identified distinct molecular subtypes of DN with different lipid metabolic characteristics. Critically, bioinformatic predictions were validated in db/db mice, showing significantly elevated mRNA and protein levels of all five biomarkers in DN kidneys.
Conclusion: Our findings reveal a validated lipid-metabolism-based diagnostic signature with potential for early detection, molecular stratification, and mechanistic investigation.

Keywords:  Diabetic nephropathy; Immune infiltration; Lipid metabolism; Machine learning; Molecular heterogeneity

DOI:  https://doi.org/10.22038/ijbms.2026.94356.20339