bims-aukdir Biomed News
on Automated knowledge discovery in diabetes research
Issue of 2026–02–15
twenty-one papers selected by
Mott Given
  1. Development and Validation of a Predictive AI Framework for Diabetic Foot Ulcer Monitoring and Severity Assessment: A Step towards Self-monitoring and Primary Care Integration.
  2. Enhanced diabetes prediction using pre-trained CNNs, LSTM, and conditional GAN on transformed numerical data.
  3. Machine learning model based on routine blood and biochemical parameters for early diagnosis of diabetic kidney disease.
  4. Machine learning based identification of key predictors and socioeconomic inequalities in the co-existence of diabetes and hypertension among Bangladeshi adults.
  5. Feasibility of Integrating Urinary Proteomics and Machine Learning for Diagnosing Diabetic Nephropathy.
  6. Diabetic Retinopathy Screening in Primary Care Real Practice: Study Procedures and Baseline Characteristics from the RETINAvalid Project.
  7. DiabLLM: An LLM-Based Framework for Blood Glucose Prediction in Type 1 Diabetes.
  8. From Telemedicine to Early Warning: Embedding Artificial Intelligence in Remote Follow-Up of Diabetic Foot Ulcers.
  9. A predictive study of glycaemic reversal in Chinese individuals with prediabetes based on machine learning: a 5-year cohort study.
  10. AFTS: A patient-agnostic encoder-decoder architecture with directional attention for blood glucose forecasting.
  11. Personalized Diabetes Treatment Support Using Large Language Models Fine-Tuned on Electronic Health Records: Development and Evaluation Study.
  12. Exploring the gut microbiome in type 2 diabetes across different insulin resistance levels: a machine learning approach.
  13. Comparison of Artificial Intelligence and Guidelines in Answering Questions on Gestational Diabetes: a CLEAR Tool Analysis.
  14. External validation of GDM risk prediction models using a machine learning reciprocal model-exchange framework.
  15. Identification of disulfidptosis-related genes in immunity and immunotherapy in diabetic foot ulcer.
  16. Identification of VEGFB associated with NKT cells in diabetic foot ulcers: Single-cell analysis and machine learning.
  17. Correction: A personalized federated learning-based glucose prediction algorithm for high-risk glycemic excursion regions in type 1 diabetes.
  18. Type 5 diabetes mellitus: redefining pancreatogenic diabetes through molecular, imaging, and AI-driven evidence.
  19. Salivary extracellular vesicles and Raman spectroscopy in precision diagnostics of type 2 diabetes.
  20. Macrovascular complications in type 2 diabetes: a multiregional study in rural Bangladesh.
  21. Prediction of stroke events in patients with type 2 diabetes mellitus by interpretable machine learning based on perivascular adipose tissue features: a multicenter cohort study.
  1. Healthc Inform Res. 2026 Jan;32(1): 69-76
    Development and Validation of a Predictive AI Framework for Diabetic Foot Ulcer Monitoring and Severity Assessment: A Step towards Self-monitoring and Primary Care Integration.
    Subodh S Satheesh, Akhila Rayampalli, Akash G Prabhune, Vinay R Sri Hari.
       OBJECTIVES: Diabetic foot ulcer (DFU) is a critical complication of diabetes that can lead to severe outcomes such as infection, amputation, and increased mortality if left untreated. Early detection and continuous monitoring are essential but remain challenging, especially in resource-limited settings such as India. This study developed and validated a deep learning algorithm to classify diabetic foot images into severity grades based on the International Working Group on the Diabetic Foot classification: grade 0 (healthy), grade 1 (mild), grade 2 (moderate), and grade 3 (severe).
    METHODS: A dataset of 407 clinical images was collected from open-source platforms and clinics in South India and expanded to 612 images through data augmentation. The dataset was divided into training (70%), validation (15%), and testing (15%) subsets. Multiple machine learning models were tested, including MobileNet_V2, EfficientNet-b0, DenseNet121, ResNet_50, VGG16, and ViT_b_16.
    RESULTS: Among the evaluated models, MobileNet_V2 demonstrated the highest validation accuracy (82%) and achieved an F1-score of 79% on the test set. Although the model showed strong training accuracy, minor overfitting was observed, particularly in distinguishing adjacent severity grades. To address this, dropout, batch normalization, and early stopping were employed. Overall, the model generalized well, showing high accuracy in detecting healthy cases and acceptable performance across ulcer severity grades.
    CONCLUSIONS: This study underscores the potential of machine learning-based tools to support frontline healthcare workers and facilitate patient self-monitoring in low-resource environments. Future work will focus on refining the model and integrating it into user-friendly applications.
    Keywords:  Computer-Assisted; Deep Learning; Diabetes Complications; Diabetic Foot; Image Processing; Machine Learning Algorithms
    DOI:  https://doi.org/10.4258/hir.2026.32.1.69
  2. Sci Rep. 2026 Feb 10.
    Enhanced diabetes prediction using pre-trained CNNs, LSTM, and conditional GAN on transformed numerical data.
    K Rupabanta Singh, Sujata Dash, Haipeng Liu, Zhen Wang.
      Diabetes remains a major public health challenge, contributing to complications such as kidney disease, cardiovascular disorders, and diabetic retinopathy. Early detection is essential for timely intervention, yet prediction from structured biomedical data is often hindered by limited sample size and feature diversity. This study investigates a deep learning framework that combines tabular-to-image transformation, pre-trained Convolutional Neural Networks, and Long Short-Term Memory (LSTM) networks to enhance diabetes prediction. Using the Pima Indians Diabetes Dataset, numerical features were transformed into 2D image representations based on correlation patterns and feature importance scores. Conditional Generative Adversarial Networks generated additional synthetic samples for training. Feature extraction was performed with DenseNet201, ResNet152, Xception, and EfficientNetB4, followed by classification using LSTM networks optimised via Bayesian search. In five-fold cross-validation, the deep learning pipeline achieved 94% accuracy and 98% AUC on the augmented PIMA dataset, showing improved performance compared to commonly reported benchmarks; however, these results may partially reflect the influence of synthetic data. When evaluated on the Frankfurt Diabetes Dataset, the model exhibited comparable performance, although the limited number of samples indicates that additional studies are required to firmly establish its generalizability. The proposed framework demonstrates promising performance for diabetes prediction from structured data. While the results suggest potential applicability to broader biomedical classification tasks, further validation on large, demographically diverse, and multi-institutional datasets is essential before considering any clinical translation.
    Keywords:  DenseNet201; EfficientNetB4; Generative adversarial network (GAN); LSTM; Pre-trained convolutional neural network; ResNet152; Type 2 diabetes; Xception
    DOI:  https://doi.org/10.1038/s41598-026-38942-5
  3. Front Endocrinol (Lausanne). 2026 ;17 1720574
    Machine learning model based on routine blood and biochemical parameters for early diagnosis of diabetic kidney disease.
    Wei Yong, Dan-Dan Peng, Kai Ye, Jun-Jie Gao, Ruo-Xue Cao.
       Background: Diabetic kidney disease (DKD) is the leading cause of end-stage renal disease globally, yet early diagnosis remains challenging due to conventional biomarker limitations, including UACR variability and reduced eGFR sensitivity. While machine learning shows promise in diabetes prediction, its application to early DKD identification using routine parameters remains underexplored. This study aimed to develop and validate machine learning models incorporating routine blood and biochemical parameters for early DKD prediction.
    Methods: This retrospective study analyzed 3,114 diabetic patients from the Second Affiliated Hospital of Wannan Medical College (EDN1) and 1,496 patients from NHANES 2005-2018 (EDN2) for external validation. Early DKD was defined as UACR 30-300 mg/g with eGFR ≥60 ml/min/1.73m². Seven machine learning algorithms were compared. Feature importance was assessed using SHAP framework, and Mendelian randomization explored causal relationships.
    Results: Among 3,114 patients, 1,333 (42.8%) had early DKD. Logistic regression achieved optimal performance (AUC = 0.689, sensitivity=40.5%, specificity=81.3%). Top predictors included triglyceride-glucose index (TyG), gender, creatinine, globulin, and age. External validation confirmed significant associations for HbA1c, globulin, TyG, and neutrophil-to-albumin ratio.
    Conclusions: The machine learning model successfully identified early DKD using routine parameters, with TyG index, HbA1c, and globulin as key predictors, demonstrating potential as a cost-effective screening tool.
    Keywords:  diabetic kidney disease; early diagnosis; machine learning; risk prediction; routine blood parameters
    DOI:  https://doi.org/10.3389/fendo.2026.1720574
  4. Sci Rep. 2026 Feb 11.
    Machine learning based identification of key predictors and socioeconomic inequalities in the co-existence of diabetes and hypertension among Bangladeshi adults.
    Rafayet Rahman Ridoy, Md Mothashin, Md Aslam Hossain, Tahira Mahbub, Sajal Chandra Barmon, Parth Kumar Saha, Abu Sayed Md Al Mamun, Md Golam Hossain.
      
    Keywords:  Coexisting; Diabetes; Hypertension; Machine learning, decomposition analysis
    DOI:  https://doi.org/10.1038/s41598-026-37966-1
  5. J Proteome Res. 2026 Feb 13.
    Feasibility of Integrating Urinary Proteomics and Machine Learning for Diagnosing Diabetic Nephropathy.
    Jiangen Yu, Di Zhou, Da Li, Yubo Chen, Dan Zhao, Fangfang Chen, Dezhen Wang, Xiuhong Li, Junli Gao, Jun Chen.
      Diabetic nephropathy (DN) represents the predominant microvascular complication associated with diabetes mellitus; however, existing diagnostic techniques are inadequate. This study evaluated candidate urinary protein biomarkers for diagnosing DN. A cohort comprising 59 patients with type 2 diabetes, 60 patients with DN, and 60 healthy volunteers was recruited. Urine proteomics was utilized to investigate differential protein expression levels among various patient groups and to identify potential biomarkers in conjunction with data analysis from the gene expression omnibus database. Machine learning classification methods were utilized to construct differential diagnosis models for DN. The data set IPX0003092000 was used to validate these diagnostic models. Six potential biomarkers─SERPINF1, FABP4, CP, CFB, C4A, and A1BG─were identified. The diagnostic models for DN, constructed by using machine learning algorithms, demonstrated robust diagnostic performance. Notably, models employing the glmnet, plr, and ranger classification methods achieved AUC values exceeding 0.800 in both the training and test data sets. In the validation cohort, the AUC values for models constructed using the ranger, glmnet, and plr methods were 0.928, 0.942, and 0.850, respectively. We evaluated six candidate urinary biomarkers (SERPINF1, FABP4, CP, CFB, C4A, and A1BG) using urinary proteomics and developed a diagnostic model for DN using machine learning algorithms.
    Keywords:  diabetic nephropathy; diagnostic model; machine learning algorithms; proteomic; urinary biomarkers
    DOI:  https://doi.org/10.1021/acs.jproteome.5c00393
  6. Healthcare (Basel). 2026 Jan 28. pii: 334. [Epub ahead of print]14(3):
    Diabetic Retinopathy Screening in Primary Care Real Practice: Study Procedures and Baseline Characteristics from the RETINAvalid Project.
    Víctor-Miguel López-Lifante, Maria Palau-Antoja, Noemí Lamonja-Vicente, Cecilia Herrero-Alonso, Josefina Sala-Leal, Rosa García-Sierra, Adrià Prior-Rovira, Marina Alventosa-Zaidin, Meritxell Carmona-Cervelló, Erik Isusquiza Garcia, Idoia Besada, Pere Torán-Monserrat.
      Background/Objectives: With rising diabetes rates, early detection of complications such as diabetic retinopathy (DR), a leading cause of visual impairment, is crucial. Incorporating DR screening into primary care has shown positive results, and integrating technological advances and artificial intelligence (AI) into these processes offers promising potential. The overall study aims to evaluate the agreement between primary care physicians, ophthalmologists, and an AI system in DR screening and referral decisions within a real-world primary care setting. Methods: In this brief report, we present the study protocol and provide an initial overview and description of our sample. A total of 1517 retinographies, obtained by a non-mydriatic retinal camera, were retrospectively collected from 301 patients with diabetes. Results: Primary care physicians referred 34.5% of the patients to ophthalmology, primarily due to opacification, suspicion of DR, or other retinal diseases. Overall, 13.62% of the participants were suspected of having DR, with 9.63% having a definitive diagnosis. Conclusions: These initial descriptive findings will be further explored in the next phase of the study through the analysis of concordance between primary care physicians, the AI-based software, and ophthalmology specialists. Future results are expected to provide valuable insights into the reliability of DR screening across different evaluators and support the integration of effective DR screening strategies into real-world clinical practice.
    Keywords:  artificial intelligence; diabetic retinopathy; primary care; retinography; screening
    DOI:  https://doi.org/10.3390/healthcare14030334
  7. IEEE J Biomed Health Inform. 2026 Feb 09. PP
    DiabLLM: An LLM-Based Framework for Blood Glucose Prediction in Type 1 Diabetes.
    Amirhossein Mahmoudi, Ghazal Farahani, Peter Domanski, Bahar Farahani, Farshad Firouzi, Krishnendu Chakrabarty.
      Accurate Blood Glucose (BG) prediction is essential for enabling glycemic control in individuals with Type 1 Diabetes Mellitus (T1DM), particularly within Smart and Connected Health (SCH) systems that integrate Continuous Glucose Monitoring (CGM) and automated insulin delivery. The adaptability of Large Language Models (LLMs) provides a promising foundation for unified, fine-tunable forecasting models. We introduce DiabLLM, a framework based on two recent LLM-based architectures: Time-LLM, which incorporates a lightweight projection layer and alignment techniques to transform time-series data into embeddings interpretable by pre-trained LLMs, and Chronos, which employs time-series-aware tokenization and quantization to convert continuous inputs into discrete sequences for forecasting. Both models process 30-minute sequences of six historical BG values and predict 30- and 45-minute horizons. Experimental results on the OhioT1DM and D1NAMO datasets demonstrate that DiabLLM outper forms state-of-the-art baselines, including a Deep Reinforcement Learning model and an ensemble of LSTM, GRU, and WaveNet, achieving up to 27% improvement in RMSE and 37% in MAE. To enhance robustness to noisy and missing input data, a denoising autoencoder was employed for input reconstruction, yielding improved predictive performance. In addition, knowledge distillation was shown to significantly compress the model, making it a practical candidate for efficient deployment on resource-constrained edge devices without compromising accuracy.
    DOI:  https://doi.org/10.1109/JBHI.2026.3658588
  8. J Diabetes Sci Technol. 2026 Feb 12. 19322968261423200
    From Telemedicine to Early Warning: Embedding Artificial Intelligence in Remote Follow-Up of Diabetic Foot Ulcers.
    Dured Dardari.
      
    Keywords:  artificial intelligence; diabetic foot ulcer; remote monitoring; telemedicine; triage
    DOI:  https://doi.org/10.1177/19322968261423200
  9. Front Endocrinol (Lausanne). 2026 ;17 1686082
    A predictive study of glycaemic reversal in Chinese individuals with prediabetes based on machine learning: a 5-year cohort study.
    Changshun Yan, Su Hu, Hangyu Cao, Rui Xu, Guiqiu Cao, Genshan Ma.
       Objective: Diabetes mellitus (DM) poses a major global public health challenge. Prediabetes, a critical stage in the progression of DM, represents a pivotal window for intervention and prevention. This study aims to develop and validate a machine learning-based prediction model for glycemic reversal in Chinese individuals with prediabetes, with the goal of facilitating such reversal in this population.
    Methods: This study analyzed data of Chinese adults from the Dryad database, with a follow-up period from 2010 to 2016. LASSO regression was used to select variables. The selected variables were then used to construct models using random forest, gradient boosting decision tree, eXtreme gradient boosting, Naive Bayes, adaptive boosting, support vector machine (SVM), and Cox model. To assess the discriminative ability of each model, the area under the curve (AUC) was calculated for each. Predictive performance was evaluated by computing time-dependent AUC (t-AUC), accuracy, precision, recall, F1, and C-index. Shapley additive explanations (SHAP) analysis was applied to interpret the key variables identified by the optimal model, and Kaplan-Meier curves for key variables associated with glycemic improvement were plotted to explore differences between groups.
    Results: 1792 adults with prediabetes were enrolled. During 5 years of follow-up, 942 achieved normoglycemia, yielding a reversal rate of 52.6%. After differential analysis and LASSO regression screening, 12 feature variables were finally determined for model construction. The 3-year, 4-year, and 5-year AUC values for the Cox model all exceeded 0.61. Six machine learning algorithms were employed to construct predictive models. The SVM demonstrated superior overall performance: it yielded a t-AUC of 0.711, accuracy of 0.652, precision of 0.620, recall of 0.661, F1 of 0.639, and a C-Index of 0.709, outperforming the other algorithms. SHAP analysis revealed that age, FPG, BMI, SBP, DBP, and triglycerides are key factors influencing normoglycemia reversal in individuals with prediabetes.
    Conclusion: We developed an SVM model to predict glycemic reversal in the prediabetic population in China, and identified key factors influencing glycemic improvement. This work provides a scientific basis for both this population and clinicians to implement early targeted interventions, thereby aiding in reducing the incidence of DM and alleviating the healthcare burden.
    Keywords:  Chinese population; diabetes mellitus; machine learning; prediabetes; prediction models
    DOI:  https://doi.org/10.3389/fendo.2026.1686082
  10. J Biomed Inform. 2026 Feb 11. pii: S1532-0464(26)00022-5. [Epub ahead of print] 104998
    AFTS: A patient-agnostic encoder-decoder architecture with directional attention for blood glucose forecasting.
    Yu Chen, Henghong Lin, Zhijin Wang, Jinmo Tang, Yaohui Huang, Xiufeng Liu, Senzhen Wu.
      Accurate blood glucose forecasting remains challenging due to inter-patient heterogeneity and complex glycemic dynamics. We present AFTS (Adaptive Feature Time Series), a patient-agnostic deep learning architecture combining a bidirectional LSTM encoder-decoder with cascaded Directional Representation (DR) modules. These modules introduce a specialized axis-wise attention mechanism that processes temporal and feature dimensions separately, designed to disentangle trend evolution from latent feature magnitude. We evaluated AFTS on two real-world CGM datasets (KDD18 and CDD23) against twenty baseline models, including advanced Transformers and RNN variants. Under a rigorous patient-wise 80/20 split, AFTS achieved an MAE of 7.02 mg/dL (KDD18) and 7.39 mg/dL (CDD23) at a 30-minute prediction horizon. The results demonstrate that AFTS is numerically competitive with state-of-the-art architectures while offering a distinct mechanism for hierarchical feature refinement. By isolating the encoder-decoder backbone and DR modules in ablation studies, we confirm that the axis-wise attention mechanism contributes specifically to minimizing prediction error in complex glycemic scenarios. These findings establish AFTS as a robust architectural candidate for patient-agnostic forecasting, effectively balancing the capture of short-term fluctuations and long-term trends.
    Keywords:  Adaptive feature time series; Blood glucose; Continuous glucose monitoring; Deep learning; Diabetes management; Patient-agnostic; Prediction
    DOI:  https://doi.org/10.1016/j.jbi.2026.104998
  11. JMIR Form Res. 2026 Feb 09. 10 e71541
    Personalized Diabetes Treatment Support Using Large Language Models Fine-Tuned on Electronic Health Records: Development and Evaluation Study.
    Shenyang He, Yu Zhang, Jiaxi Li.
       Background: Effective diabetes management requires individualized treatment strategies tailored to patients' clinical characteristics. With recent advances in artificial intelligence, large language models (LLMs) offer new opportunities to enhance clinical decision support, particularly in generating personalized recommendations.
    Objective: This study aimed to develop and evaluate an LLM-based outpatient treatment support system for diabetes and examine its potential value in routine clinical decision-making.
    Methods: Three compact LLMs (Llama 3.1-8B, Qwen3-8B, and GLM4-9B) were fine-tuned on deidentified outpatient electronic health records using a parameter-efficient low-rank adaptation approach. The optimized models were embedded into a prototype hospital information system via a retrieval-augmented generation framework to generate individualized treatment recommendations, laboratory test suggestions, and medication prompts based on demographic and clinical data.
    Results: Among the models evaluated, the fine-tuned GLM4-9B demonstrated the strongest performance, producing clinically reasonable treatment plans and appropriate laboratory test recommendations and medication suggestions. It achieved a mean Bilingual Evaluation Understudy for 4-grams score of 67.93 (SD 2.74) and mean scores of 44.30 (SD 3.91) for Recall-Oriented Understudy for Gisting Evaluation for overlap of unigrams, 27.34 (SD 1.85) for Recall-Oriented Understudy for Gisting Evaluation for overlap of bigrams, and 37.67 (SD 2.88) for Recall-Oriented Understudy for Gisting Evaluation for Longest Common Subsequence.
    Conclusions: The fine-tuned GLM4-9B shows strong potential as a clinical decision support tool for personalized diabetes care. It can provide reference recommendations that may improve clinician efficiency and support decision quality. Future work should focus on enhancing medication guidance, expanding data sources, and improving adaptability in cases involving complex comorbidities.
    Keywords:  AI; EHR; GLM4-9B; artificial intelligence; diabetes; electronic health record; large language model
    DOI:  https://doi.org/10.2196/71541
  12. Front Nutr. 2026 ;13 1747767
    Exploring the gut microbiome in type 2 diabetes across different insulin resistance levels: a machine learning approach.
    Yuchi He, Lu Liu, Yifan Liu, Jialong Jia, Yuqing Chen, Xiyu Zhang, Ya Liu.
       Introduction: Insulin resistance (IR) is central to type 2 diabetes mellitus (T2DM). Composite indices including the atherogenic index of plasma (AIP), metabolic score for insulin resistance (METS-IR), triglyceride-glucose index (TyG), and TyG-BMI, are widely used to quantify IR severity. The gut microbiome (GM) has been implicated in metabolic dysregulation, but its associations with IR remain incompletely defined.
    Methods: We collected blood test results and stool samples from participants with T2DM and healthy controls. Stool samples underwent 16S rRNA gene sequencing. We trained XGBoost models to distinguish individuals with higher IR from healthy controls based on GM profiles and performed correlation analyses between GM features, clinical measures, and IR indices.
    Results: Triglycerides (TG), fasting blood glucose (FBG), and high-density lipoprotein cholesterol (HDL-C) differed significantly between the T2DM and control groups. IR indices (AIP, METS-IR, TyG, and TyG-BMI) were markedly higher in the T2DM group. XGBoost models based on GM profiles showed high discriminatory performance for identifying T2DM individuals with higher IR, with Bacteroides and Faecalibacterium contributing most to model performance. Correlation analyses further indicated that Lachnospiraceae_UCG-010, Bacteroides, Faecalibacterium, Lachnospira, Parasutterella, and Escherichia-Shigella were associated with clinical measures and IR indices.
    Conclusions: Specific GM features are associated with IR-related clinical measures and composite indices in T2DM, supporting their potential as intervention targets to improve insulin resistance and restore carbohydrate and lipid metabolism.
    Keywords:  XGBoost; gut microbiome; insulin resistance; machine learning; type 2 diabetes mellitus
    DOI:  https://doi.org/10.3389/fnut.2026.1747767
  13. Clin Lab. 2026 Feb 01. 72(2):
    Comparison of Artificial Intelligence and Guidelines in Answering Questions on Gestational Diabetes: a CLEAR Tool Analysis.
    Gökhan Köker, Lütfullah Z Koç, Muhammed A Coşkuner, Yasin Şahintürk, Bilgin B Başgöz.
       BACKGROUND: Gestational diabetes mellitus (GDM) affects millions of people worldwide. Patients often turn to the internet and artificial intelligence (AI)-based conversational models for information. The CLEAR tool evaluates the quality of health-related content produced by AI-based models. This study assessed the responses provided by medical guidelines, ChatGPT, and Google Bard to the ten most frequently asked online questions about GDM, uti-lizing the CLEAR tool for evaluation.
    METHODS: The most common online questions about GDM were identified using Google Trends, and the top 10 questions were selected. Answers were then gathered from two experienced physicians, ChatGPT 4.0o-mini, and Google Bard, with responses categorized into 'Guide,' 'ChatGPT,' and 'Bard' groups. Answers from the AI models were obtained using two computers and two separate sessions to ensure consistency and minimize bias.
    RESULTS: ChatGPT received higher scores than the medical guidelines, while Bard scored lower than ChatGPT. The medical guidelines provided more accessible answers for the general audience, while ChatGPT and Bard required higher literacy levels. Good reliability (0.781) was observed between the two reviewers. Regarding readability, the medical guidelines were the easiest to read, while Bard provided the most challenging text.
    CONCLUSIONS: ChatGPT and Google Bard perform well in content completeness and relevance but face challenges in readability and misinformation. Future research should improve accuracy and readability, integrate AI with peer-reviewed sources, and ensure healthcare professionals guide patients to reliable AI information.
    DOI:  https://doi.org/10.7754/Clin.Lab.2025.250544
  14. Comput Biol Med. 2026 Feb 11. pii: S0010-4825(26)00109-5. [Epub ahead of print]204 111547
    External validation of GDM risk prediction models using a machine learning reciprocal model-exchange framework.
    Mark Germaine, Yitayeh Belsti, Amy O'Higgins, Brendan Egan, Helena Teede, Graham Healy, Joanne Enticott.
       BACKGROUND: Although many risk prediction models have been developed, very few undergo external validation, primarily due to issues with data access. Therefore, we implemented a reciprocal model-exchange approach to facilitate external validation and demonstrate its use with gestational diabetes mellitus (GDM) prediction models.
    OBJECTIVE: To assess the robustness and generalisability of two independently developed GDM risk prediction models using a reciprocal model-exchange framework.
    METHODS: Two independently developed GDM risk prediction models were externally validated using a reciprocal model-exchange. The saved model's corresponding variable types and data pre-processor were exchanged. The Monash CatBoost model was validated using Irish data at Dublin City University (DCU), and the DCU logistic-regression GDM model was validated using Australian data at Monash University. Performance was assessed using discrimination, calibration and decision curve analysis. Model fairness was assessed.
    RESULTS: The prevalence of GDM was 21.1% in the Australian cohort and 11.7% in the Irish cohort. The Monash model's AUC dropped from 0.93 to 0.77, while the DCU model's AUC fell from 0.82 to 0.69. Calibration estimates confirmed systematic risk misestimation; each model tends to over or under-predict GDM probabilities outside its training domain, with calibration-in-the-large of -0.573 for the Monash model and 0.17 for the DCU model; slopes were 1.278 and 0.55 respectively. Both models showed performance variability across ethnic groups, with lower performance for Southeast/Northeast Asians and both performed better with increasing parity and among women without a prior GDM diagnosis.
    CONCLUSIONS: Each model's performance decreased upon external validation, and the fairness evaluations on the different sub-categories (ethnicities; parity and previous GDM) provided evidence on the areas to be addressed in model recalibration/updating before deployment can be progressed. This reciprocal model-exchange approach provides a solution to facilitating external validations, which are notably lacking in the current literature but are necessary to advance the risk prediction field.
    DOI:  https://doi.org/10.1016/j.compbiomed.2026.111547
  15. Ann Med Surg (Lond). 2026 Feb;88(2): 1402-1414
    Identification of disulfidptosis-related genes in immunity and immunotherapy in diabetic foot ulcer.
    JIe Li, Hongshuo Shi, Yemin Cao.
       Background: Diabetic foot ulcer (DFU) is one of the most common and severe complications of diabetes, with vascular changes, neuropathy, and infections being the primary pathological mechanisms. Disulfidptosis, a recently identified form of programmed cell death, might be involved in the development of diabetic complications. This study aims to identify and validate potential disulfidptosis biomarkers associated with DFU through bioinformatics and machine learning analysis.
    Methods: We downloaded two microarray datasets related to DFU patients from the Gene Expression Omnibus (GEO) database, namely GSE134431, GSE68183, and GSE80178. From the GSE134431 dataset, we obtained differentially expressed Gln-metabolism-related genes (deDRGs) between DFU and normal controls. We analyzed the correlation between deDRGs and immune cell infiltration status. We also explored the relationship between DRG molecular clusters and immune cell infiltration status. Notably, we used Weighted Gene Co-expression Network Analysis (WGCNA) to identify differentially expressed genes within specific clusters. We used Gene Set Variation Analysis (GSVA) to explore which pathways might be related to the DRGs. Subsequently, we constructed and screened the best machine learning model. Finally, we validated the predictions' accuracy using a nomogram, calibration curves, decision curve analysis, and the GSE80178 and GSE68183 datasets.
    Results: In both the DFU and normal control groups, we confirmed the presence of deDRGs and an activated immune response. From the GSE134431 dataset, we obtained 33 deDRGs, including MYH10, MYL6, UBASH3B, SLC7A11, DSTN, CD2AP, ME1, OXSM, NDUFC1, GYS1, SCO2, NLN, HNRNPH2, MRPS17, SART3, SAFB2, SAFB, HNRNPU, HNRNPM, MYH14, GTF2I, MYH3, CNOT1, PCBP2, GLUD1, MYH11, TLN2, CHD4, SQSTM1, NDUFB11, NDUFS2, SAMM50, and PPIH. Furthermore, two clusters were identified in DFU. Immune infiltration analysis indicated the presence of immune heterogeneity in these two clusters. Additionally, we established a support vector machine model based on five genes (RALY, R3HCC1, CES1, TCEAL3, and F13A1), which exhibited excellent performance on the external validation datasets GSE80178 and GSE68183 (AUC = 1).
    Conclusion: This study has identified five disulfidptosis genes associated with DFU, revealing potential novel biomarkers and therapeutic targets for DFU. Additionally, the infiltration of immune-inflammatory cells plays a crucial role in the progression of DFU.
    Keywords:  diabetic foot ulcer; disulfidptosis; immune infiltration; machine learning; molecular clusters
    DOI:  https://doi.org/10.1097/MS9.0000000000003859
  16. Medicine (Baltimore). 2026 Feb 13. 105(7): e47574
    Identification of VEGFB associated with NKT cells in diabetic foot ulcers: Single-cell analysis and machine learning.
    Weilun Wang, Yao He, Tao You, Meihong Lei, Jianming Chen.
      The management options for diabetic foot are restricted, and the outlook is unfavorable. Immune cells have been implicated in diabetic foot ulcer (DFU), but the exact role of natural killer T (NKT) cells in DFU remains unclear. Vascular endothelial growth factor B (VEGFB), a member of the VEGF family, is distinguished by its potential roles in metabolic regulation and immune modulation, yet its connection to NKT cells in DFU is unexplored. This study was to identify specific genes associated with NKT cells in DFU and to ascertain potential targets. We analyzed single-cell ribonucleic acid sequencing and bulk transcriptome data from DFU datasets. Differential expression analysis identified genes associated with NKT cells in DFU. Machine learning algorithms were applied to pinpoint the most significant genes from these candidates. The functional characteristics of the identified key gene were further investigated through gene set enrichment analysis and immune infiltration analysis. Single-cell analysis revealed 390 NKT cell-related genes, and differential analysis identified 728 differentially expressed genes. Cross-referencing yielded 37 NKT cell-related differentially expressed genes. Machine learning consistently identified VEGFB as a key biomarker. Functional analysis linked VEGFB to cell adhesion, vasculature development, and angiogenesis pathways. VEGFB was significantly overexpressed in DFU samples compared to controls. Our study identifies VEGFB as a valuable biomarker associated with NKT cells in DFU. The overexpression of VEGFB suggests its involvement in DFU pathogenesis, potentially bridging immune regulation and vascular pathways. This finding enhances the understanding of NKT cell mechanisms in DFU and positions VEGFB as a potential target for future diagnostic and therapeutic strategies aimed at immunomodulation.
    Keywords:  NKT cell; diabetes foot ulcers; machine learning; single cell analysis
    DOI:  https://doi.org/10.1097/MD.0000000000047574
  17. Sci Rep. 2026 Feb 10. 16(1): 5742
    Correction: A personalized federated learning-based glucose prediction algorithm for high-risk glycemic excursion regions in type 1 diabetes.
    Dave Darpit, Kathan Vyas, Jagadish Kumaran Jayagopal, Alfredo Garcia, Madhav Erraguntla, Mark Lawley.
      
    DOI:  https://doi.org/10.1038/s41598-026-36109-w
  18. Front Endocrinol (Lausanne). 2025 ;16 1749805
    Type 5 diabetes mellitus: redefining pancreatogenic diabetes through molecular, imaging, and AI-driven evidence.
    Imran Rashid Rangraze, Mohamed El-Tanani, Adil Farooq Wali, Rasha Babiker, Syed Arman Rabbani, Ismail I Matalka, Shakta Mani Satyam, Ashot Avagimyan, Karolina Hoffmann, Ioannis Ilias, Sorina Ispas, Maggio Viviana, Anna Paczkowska, Manfredi Rizzo.
       Background: Type 5 Diabetes Mellitus (T5DM), denoting pancreatogenic diabetes from fibro-inflammatory pancreatic injury, is a distinct yet under-recognised entity. Current WHO and ADA classifications overlook its complex, concurrent endocrine-exocrine failures, contributing to misdiagnosis, treatment gaps, and suboptimal outcomes.
    Objectives: This review aims to critically analyze current scientific understanding of the pathogenesis, diagnostic criteria, metabolic consequences, and therapeutic needs of T5DM and suggest a precise framework of medicine that justifies the need for T5DM to be formally recognized as a sub-type of diabetes.
    Methods: An integrative review was conducted using recent literature on pancreatic pathophysiology, molecular biomarkers, radiomics, diagnostic imaging, glycemic control technologies, and machine learning. The focus was on the recent literature to elucidate the biological, diagnostic, and treatment aspects of the clinical studies, guidelines, and mechanistic research available from the publications.
    Key findings: T5DM involves loss of insulin and glucagon alongside exocrine pancreatic insufficiency, malnutrition, and significant glycaemic variability. A tiered diagnostic framework-integrating pancreatic imaging, endocrine-exocrine testing, autoimmune exclusion, and emerging biomarkers-enhances accuracy. Management requires coordinated hormonal and enzyme replacement, structured nutritional support, and targeted surveillance for malignancy and micronutrient deficits. Radiomics, quantitative imaging, and AI-driven analytics offer valuable tools for earlier detection, improved risk stratification, and personalised therapy.
    Conclusion: T5DM warrants recognition as a distinct diabetes entity owing to its unique pathophysiology, clinical behaviour, and therapeutic needs. Harmonised diagnostic criteria, validated biomarker and imaging pathways, and multicentre registries are essential to integrate T5DM into global classification systems and advance mechanism-based, personalised care.
    Keywords:  AI diagnostics; biomarkers; exocrine pancreatic insufficiency; pancreatogenic diabetes; precision medicine; reclassification; type 5 diabetes mellitus
    DOI:  https://doi.org/10.3389/fendo.2025.1749805
  19. Diabetes Res Clin Pract. 2026 Feb 05. pii: S0168-8227(26)00058-6. [Epub ahead of print]233 113139
    Salivary extracellular vesicles and Raman spectroscopy in precision diagnostics of type 2 diabetes.
    Ajitesh Dhal, Shao-Jung Lin, Arunima Pandey, Chih-Hsuan Liu, Hung-Yi Liu, Tarakanta Jena, Chitralekha Jena, Dharitri Rath, Pei-Wen Peng, Cheng-Jen Chang, Chang-I Chen, Li-Chern Pan, Tzu-Sen Yang.
      Type 2 Diabetes (T2D) remains a major global health issue, driven by sedentary lifestyles and aging populations, emphasizing the urgent need for precise diagnostics that allow early detection and personalized monitoring. Traditional blood tests, including glucose and HbA1c measurements, offer limited temporal and molecular information. In contrast, saliva provides a non-invasive, easily accessible biofluid that reflects systemic metabolic changes. Its molecular components, especially extracellular vesicles (EVs), such as exosomes and microvesicles, contain proteins, lipids, and microRNAs directly associated with insulin resistance, β-cell dysfunction, and inflammation in T2D. Advances in Raman spectroscopy and surface-enhanced Raman scattering (SERS) now enable high-sensitivity, label-free molecular fingerprinting of salivary EVs, supporting multiplex detection of disease-related biomarkers. Combining Raman-based sensing with EV profiling introduces an innovative approach for non-invasive, precision diabetes diagnostics. This review explores the diagnostic importance of salivary EVs, recent developments in Raman/SERS-based biomolecular detection, and the clinical potential of integrating these technologies for early screening and therapy monitoring. Moreover, incorporating artificial intelligence (AI) for spectral analysis and developing portable Raman devices could facilitate real-time, saliva-based metabolic monitoring, advancing personalized, preventive, and patient-focused diabetes care.
    Keywords:  Artificial intelligence (AI); Extracellular vesicles (EVs); Raman spectroscopy; Saliva; Type 2 diabetes (T2D)
    DOI:  https://doi.org/10.1016/j.diabres.2026.113139
  20. Front Endocrinol (Lausanne). 2026 ;17 1724957
    Macrovascular complications in type 2 diabetes: a multiregional study in rural Bangladesh.
    Bodrun Naher Siddiquea, Dianna J Magliano, Maima Matin, Afsana Afroz, Baki Billah.
       Objectives: To assess the prevalence and determinants of macrovascular complications (coronary artery disease, stroke, and diabetic foot) among adults living with T2DM in rural Bangladesh.
    Methods: A population-based cross-sectional study was conducted between December 2023 and September 2024, involving 1094 adults with diagnosed T2DM from rural areas of three regions/divisions in Bangladesh. Data were collected through household interviews, physical examination, and medical record reviews. Macrovascular complications were identified using clinical criteria and documented diagnosis. The leverage of six machine learning (ML) algorithms were applied in identifying influential variables associated with these complications.
    Results: The prevalence of coronary artery disease (CAD), stroke, and diabetic foot was 11.2%, 5.3%, and 9.1%, respectively. The Light Gradient Boosting Machine algorithm performed best for CAD and diabetic foot, with ROC values of 98.8% and 92.6%, respectively, while Random Forest showed the best performance for stroke with a ROC of 99%. These models also outperformed others across accuracy, precision, F1 score, and calibration. Across models, common predictors included older age, longer diabetes duration, diabetes onset at age 45 years or above, and smoking. Hypertension and elevated cholesterol were linked to CAD and stroke. Coexisting microvascular complications were also identified.
    Conclusions: This study identified a substantial burden of macrovascular complications among rural adults with T2DM, with CAD, stroke, and diabetic foot emerging as the most prevalent outcomes. Advanced age, longer duration of diabetes, smoking, hypertension, and elevated cholesterol were consistently associated with these complications, highlighting the need for intensified cardiometabolic risk control within primary care. These findings underscore the urgency of strengthening integrated diabetes-cardiovascular management in rural Bangladesh to reduce the progression and impact of these major vascular outcomes.
    Keywords:  Bangladesh; Type 2 diabetes; coronary artery disease; diabetic foot; machine learning algorithm; macrovascular complications; stroke
    DOI:  https://doi.org/10.3389/fendo.2026.1724957
  21. Quant Imaging Med Surg. 2026 Feb 01. 16(2): 114
    Prediction of stroke events in patients with type 2 diabetes mellitus by interpretable machine learning based on perivascular adipose tissue features: a multicenter cohort study.
    Ting Zhao, Guihan Lin, Weiyue Chen, Chengli Jiang, Weiming Hu, Lei Xu, Yongjun Chen, Yang Jing, Jinhong Sun, Zhihan Yan, Shuiwei Xia, Chenying Lu, Minjiang Chen, Jiansong Ji, Weiqian Chen.
       Background: Stroke is one of the leading causes of mortality, and patients with type 2 diabetes mellitus (T2DM) have a higher incidence of stroke. However, research on the imaging characteristics of plaques and perivascular adipose tissue (PVAT) in this patient population remains limited. This study therefore aimed to develop and validate a machine learning-based combined model to predict acute stroke events in patients with T2DM and assess its utility in stratifying patients into different risk categories based on follow-up outcomes.
    Methods: In this multicenter study, a total of 494 computed tomography angiography (CTA) datasets from patients with T2DM were retrospectively collected from The Fifth Affiliated Hospital of Wenzhou Medical University, The Second Affiliated Hospital of Wenzhou Medical University, and Lishui People's Hospital and divided into four sets: training (n=193), internal testing (n=84), external validation 1 (n=105), and external validation 2 (n=102). Based on the magnetic resonance imaging findings, the patients were divided into a stroke group and a non-stroke group. PVAT features were extracted from CTA, and perivascular fat density (PFD) was determined. A combined model was developed by integrating radiomics scores with PFD and clinical factors via the extreme gradient boosting (XGBoost) algorithm. The model's prediction process was illustrated with the SHapley Additive exPlanation (SHAP) method, and its prognostic value was evaluated with Kaplan-Meier analysis.
    Results: In this study, 167 patients with T2DM (33.8%) who experienced ischemic stroke (IS) were classified into the stroke group, while 327 patients with T2DM (66.2%) were classified into the non-stroke group. Through application of variance thresholding, SelectKBest, and least absolute shrinkage and selection operator, seven radiomic features were ultimately selected from CTA images to construct the radiomics model. After univariate and multivariate logistic regression analysis, total cholesterol (P=0.033) and hypertension (P=0.028) were identified as independent risk factors for IS. The combined model demonstrated substantial accuracy and robustness, with an area under the receiver operating characteristic curve of 0.955, 0.847, 0.856, and 0.876 in the training, internal testing, external validation 1, and external validation 2 cohorts. SHAP analysis revealed that Exponential_glszm_SizeZoneNonUniformity and Wavelet-HLL_firstorder_Range were the most important features. Event-free survival (EFS) analysis demonstrated that the model could effectively determine patient prognosis. Results from univariate and multivariate Cox regression analyses identified the independent prognostic predictors of follow-up ischemic events to be stroke status [hazard ratio (HR) =3.916; 95% confidence interval (CI): 1.792-6.558; P<0.001] and predicted stroke status (HR =1.352; 95% CI: 1.317-4.777; P=0.030), indicating these factors are associated with the occurrence of ischemic cerebrovascular events during follow-up.
    Conclusions: The combined XGBoost model incorporating PVAT features accurately predicted stroke events in patients with T2DM and provided risk stratification for patients.
    Keywords:  SHapley Additive exPlanation (SHAP); Type 2 diabetes mellitus (T2DM); event-free survival (EFS); ischemic stroke (IS); perivascular adipose tissue (PVAT)
    DOI:  https://doi.org/10.21037/qims-2025-1760
This page is being maintained by Thomas Krichel. It was last updated on 2026‒02‒15 at 0:14.