bims-aukdir 2025-09-28 papers

bims-aukdir

Biomed News

on Automated knowledge discovery in diabetes research

Issue of 2025–09–28
25 papers selected by
Mott Given

A Multi-Model Image Enhancement and Tailored U-Net Architecture for Robust Diabetic Retinopathy Grading.
Developing a Simple Non-Laboratory-Based Machine Learning Tool for Prediabetes Screening in a Target Population: A Proof-of-Concept Study.
Development and validation of a machine learning model for cardiovascular disease risk prediction in type 2 diabetes patients.
Appraisal of Clinical Explanatory Variables in Subtyping of Type 2 Diabetes Using Machine Learning Models.
Introduction of hierarchical deep features based on the VGG-16 model for enhancing diabetic retinopathy classification.
Advanced supervised machine learning methods for precise diabetes mellitus prediction using feature selection.
Characterizing clinical risk profiles of major complications in type 2 diabetes mellitus using deep learning algorithms.
Artificial intelligence in proliferative diabetic retinopathy: advancing diagnosis, precision surgery, and anti-VEGF therapy optimization.
Ratio of haemorrhagic area to retinal area as a novel indicator for AI-based screening of diabetic retinopathy in type 2 diabetes: a community-based cross-sectional study.
Diabetic retinopathy as the primary predictor of mild cognitive impairment in type 2 diabetes: Insights from machine learning models.
Retraction: Improved support vector machine based on CNN-SVD for vision-threatening diabetic retinopathy detection and classification.
HIRD-Net: An Explainable CNN-Based Framework with Attention Mechanism for Diabetic Retinopathy Diagnosis Using CLAHE-D-DoG Enhanced Fundus Images.
Interpreting Deep Neural Networks in Diabetic Retinopathy Grading: A Comparison with Human Decision Criteria.
GlucoNet-MM: A multimodal attention-based multi-task learning framework with decision transformer for personalised and explainable blood glucose forecasting.
Toward a Clinically Actionable, Electronic Health Record-Based Machine Learning Model to Forecast 90-Day Change in Hemoglobin A1c in Youth With Type 1 Diabetes: Feasibility and Model Development Study.
A Comparative Study Between Clinical Optical Coherence Tomography (OCT) Analysis and Artificial Intelligence-Based Quantitative Evaluation in the Diagnosis of Diabetic Macular Edema.
Precision Medicine for Diabetic Retinopathy: Integrating Genetics, Biomarkers, Lifestyle, and AI.
Diabetic Foot Ulcer Classification Models Using Artificial Intelligence and Machine Learning Techniques: Systematic Review.
Predicting the future risk of developing type 2 diabetes in women with a history of gestational diabetes mellitus using machine learning and explainable artificial intelligence.
Patient acceptance of AI-assisted diabetic retinopathy screening in primary care: findings from a questionnaire-based feasibility study.
Comprehensive Survey of OCT-Based Disorders Diagnosis: From Feature Extraction Methods to Robust Security Frameworks.
Determining the association of C-reactive-protein-triglyceride-glucose index and diabetes using machine learning and LASSO regression: A cross-sectional analysis of NHANES 2001 to 2010 results.
AI-based integration of ECG biomarkers for assessing cardiac risk in type 2 diabetes mellitus with comorbid conditions for patient stratification.
A Scoping Review of Artificial Intelligence-Based Health Education Interventions for Patients with Type 2 Diabetes.
Multimodal MRI analysis selecting key brain features for machine learning based classification of diabetic neuropathic pain and phenotypes.

Diagnostics (Basel). 2025 Sep 17. pii: 2355. [Epub ahead of print]15(18):

A Multi-Model Image Enhancement and Tailored U-Net Architecture for Robust Diabetic Retinopathy Grading.

Archana Singh, Sushma Jain, Vinay Arora.

  Background: Diabetic retinopathy (DR) is a leading cause of preventable vision impairment in individuals with diabetes. Early detection is essential, yet often hindered by subtle disease progression and reliance on manual expert screening. This study introduces an AI-based framework designed to achieve robust multiclass DR classification from retinal fundus images, addressing the challenges of early diagnosis and fine-grained lesion discrimination. Methods: The framework incorporates preprocessing steps such as pixel intensity normalization and geometric correction. A Hybrid Local-Global Retina Super-Resolution (HLG-RetinaSR) module is developed, combining deformable convolutional networks for local lesion enhancement with vision transformers for global contextual representation. Classification is performed using a hierarchical approach that integrates three models: a Convolutional Neural Network (CNN), DenseNet-121, and a custom multi-branch RefineNet-U architecture. Results: Experimental evaluation demonstrates that the combined HLG-RetinaSR and RefineNet-U approach consistently achieves precision, recall, F1-score, and accuracy values exceeding 99% across all DR severity levels. The system effectively emphasizes vascular abnormalities while suppressing background noise, surpassing existing state-of-the-art methods in accuracy and robustness. Conclusions: The proposed hybrid pipeline delivers a scalable, interpretable, and clinically relevant solution for DR screening. By improving diagnostic reliability and supporting early intervention, the system holds strong potential to assist ophthalmologists in reducing preventable vision loss.

Keywords:  U-Net; biomedical image processing; data augmentation; deep learning; diabetic retinopathy; image classification

DOI:  https://doi.org/10.3390/diagnostics15182355
J Diabetes Sci Technol. 2025 Sep 22. 19322968251376380

Developing a Simple Non-Laboratory-Based Machine Learning Tool for Prediabetes Screening in a Target Population: A Proof-of-Concept Study.

Tanja Fredensborg Holm, Thomas Kronborg, Morten Hasselstrøm Jensen, Stine Hangaard.

   BACKGROUND: Progression from prediabetes to type 2 diabetes (T2D) can be delayed with early detection and intervention. Current detection methods, relying on costly blood glucose tests, limit widespread screening. Machine learning models offer the potential for non-laboratory-based tools. However, existing prediabetes detection models lack validation in their intended target populations. Thus, this study aimed to develop and validate a non-laboratory-based machine learning tool for prediabetes detection in a specific target population.
METHODS: Based on 501 adults from a prediabetes screening project, a decision tree model was developed. Twelve potential non-laboratory-based features were extracted. The target variable was categorized into prediabetes (hemoglobin A1c [HbA1c] ≥39 mmol/mol and <48 mmol/mol) and normoglycemia (HbA1c <39 mmol/mol). The data set was divided into 70% for training and 30% for validation, and forward feature selection was used to identify the most relevant features.
RESULTS: Out of 501 participants, 88 were identified with prediabetes. The mean age and body mass index (BMI) were approximately 50 years and 27 in both the training and validation sets. Forward selection identified age and waist circumference as the most important features to include in the model. The model achieved an area under the receiver operating characteristic curve (ROC AUC) of 0.8297 and 0.7961 on the training and validation sets.
CONCLUSION: A machine learning screening tool using age and waist circumference was developed with promising results. Its simplicity, by only requiring two non-laboratory features, allows for easy implementation. However, to verify the model's generalizability and external validity, it needs to be evaluated using additional data.

Keywords:  decision support; machine learning; non-laboratory; prediabetes; screening tool; type 2 diabetes prevention

DOI:  https://doi.org/10.1177/19322968251376380
Sci Rep. 2025 Sep 25. 15(1): 32818

Development and validation of a machine learning model for cardiovascular disease risk prediction in type 2 diabetes patients.

Chunming Xu, Fachao Shi, Wenlong Ding, Cunming Fang, Caoyang Fang.

  Patients with type 2 diabetes mellitus (T2DM) have a significantly higher risk of cardiovascular disease (CVD) compared to the general population. Accurately predicting this risk is crucial for developing personalized treatment plans and public health interventions. This study aims to develop and validate a model for predicting CVD risk in T2DM patients using the Boruta feature selection algorithm and machine learning methods. We analyzed data from the National Health and Nutrition Examination Survey (NHANES) from 1999 to 2018. Six machine learning (ML) models, including Multilayer Perceptron (MLP), Light Gradient Boosting Machine (LightGBM), Decision Tree (DT), Extreme Gradient Boosting (XGBoost), Logistic Regression (LR), and k-Nearest Neighbors (KNN), were employed for model development and validation. Boruta was used for optimal feature selection. The performance of the machine learning models was comprehensively evaluated using ROC curves, accuracy, and other related metrics. Shapley Additive Explanation (SHAP) analysis was conducted for visual interpretation, and the Shinyapps.io platform was utilized to deploy the best-performing models as web-based applications. A total of 4,015 T2DM patients were included, among which 999 (24.9%) had CVD. Model evaluation revealed significant overfitting with the KNN algorithm, which showed perfect discrimination in the training set but performed poorly in the test set (AUC = 0.64). In contrast, XGBoost demonstrated more consistent performance between training and testing datasets (AUC = 0.75 and 0.72, respectively), indicating better generalization ability and making it more suitable for clinical application. Using SHAP analysis, the top 10 important influencing factors identified by the XGBoost model were utilized to construct a CVD risk prediction platform for T2DM patients. The prediction model based on Boruta feature selection and machine learning shows promising results in assessing the CVD risk among T2DM patients. This study provides a viable tool for clinical use, facilitating early intervention and precision treatment.

Keywords:  Boruta feature selection; CVD; Cardiovascular disease; Machine learning; T2DM; Type 2 diabetes

DOI:  https://doi.org/10.1038/s41598-025-18443-7
J Clin Med. 2025 Sep 17. pii: 6548. [Epub ahead of print]14(18):

Appraisal of Clinical Explanatory Variables in Subtyping of Type 2 Diabetes Using Machine Learning Models.

Amar H Khamis, Fatima Abdul, Stafny Dsouza, Fatima Sulaiman, Costerwell Khyreim, Mohammed E Siddig, Riad Bayoumi.

  Background: Clustering type 2 diabetes (T2D) remains a challenge due to its clinical heterogeneity and multifactorial nature. We aimed to evaluate the validity and robustness of the clinical variables in defining T2D subtypes using a discovery-to-prediction design. Methods: Five explanatory clinical aetiology variables (fasting serum insulin, fasting blood glucose, body mass index, age at diagnosis and HbA1c) were assessed for clustering T2D subtypes using two independent patient datasets. Clustering was performed using the IBM-Modeler Auto-Cluster. The resulting cluster validity was tested by multinomial logistic regression. The variables' validity for direct unsupervised clustering was compared with machine learning (ML) predictive models. Results: Five distinct subtypes were consistently identified: severe insulin-resistant diabetes (SIRD), severe insulin-deficient diabetes (SIDD), mild obesity-related diabetes (MOD), mild age-related diabetes (MARD), and mild early-onset diabetes (MEOD). Using all five variables yielded the highest concordance between clustering methods. Concordance was strongest for SIRD and SIDD, reflecting their distinct clinical signatures in contrast to that in MARD, MOD and MEOD. Conclusions: These findings support the robustness of clinically defined T2D subtypes and demonstrate the value of probabilistic clustering combined with ML for advancing precision diabetes care.

Keywords:  aetiology; artificial intelligence; logistic regression; subtyping; type 2 diabetes

DOI:  https://doi.org/10.3390/jcm14186548
Health Inf Sci Syst. 2025 Dec;13(1): 57

Introduction of hierarchical deep features based on the VGG-16 model for enhancing diabetic retinopathy classification.

Fatemeh Taheri, Kambiz Rahbar.

  Diabetic Retinopathy (DR) is a common ocular disease that presents a significant risk of vision loss in individuals with diabetes. Accurate DR classification is critical for preventing disease progression and preserving patients' vision. However, DR classification is often complex due to the high degree of similarity and overlap among features across its different stages. To address these challenges, this study introduces a computer-aided diagnosis framework that leverages deep neural networks to extract hierarchical features. In this framework, a hierarchical semantic resolution pyramid of retinal images is generated by integrating feature maps from the pooling layers of a deep neural network model. Handcrafted features are subsequently extracted from the HSRP and then fused to create a comprehensive feature vector. This approach fuses pre-trained neural networks with handcrafted features to effectively identify distinctive image characteristics. Notably, due to the independence of feature maps from input data, the proposed architecture does not require retraining or fine-tuning, enhancing its generalizability across different image domains. To identify the most effective discriminative features for classifying DR stages, the Boruta-Shap algorithm is applied to the feature vector. With the high discriminative power of the selected features and the limited dataset size, a random forest classifier is used to categorize DR into five stages. Comparative performance analysis with existing methods demonstrates the effectiveness of the proposed approach in overcoming the challenges associated with DR classification.

Keywords:  Deep features; Diabetic retinopathy classification; Hierarchical features; VGG-16 model

DOI:  https://doi.org/10.1007/s13755-025-00371-5
Front Med (Lausanne). 2025 ;12 1620268

Advanced supervised machine learning methods for precise diabetes mellitus prediction using feature selection.

Gufran Ahmad Ansari, Salliah Shafi, Mohd Dilshad Ansari, Azhar Shadab.

   Background: Diabetes mellitus (DM) is a chronic metabolic disorder that poses a significant global health challenge, affecting millions, many of whom remain undiagnosed in the early stages. If left untreated, diabetes can result in severe complications such as blindness, stroke, cancer, joint pain, and kidney failure. Accurate and early prediction is critical for timely intervention. Recent advancements in machine learning techniques (MLT) have shown promising potential in enhancing disease prediction due to their robust pattern recognition and classification capabilities.
Materials and methods: This study presents a comparative analysis of supervised MLT such as Support Vector Machine (SVM), Naïve Bayes (NB), K-Nearest Neighbors (KNN), and Random Forest (RF) using the Pima Indian Diabetes dataset (PIDD) from the UCI repository. A 10-fold cross-validation approach was employed to mitigate class imbalance and ensure generalizability. Performance was evaluated using standard classification metrics: accuracy, precision, recall, and F1-score.
Results: Among the evaluated models, SVM outperformed the others with an accuracy of 91.5%, followed by RF (90%), KNN (89%), and NB (83%). The study highlights the effectiveness of SVM in early diabetes prediction and demonstrates how model performance varies with algorithm selection.
Conclusion: Unlike many prior studies that focus on a single algorithm or overlook validation robustness, this research offers a comprehensive comparison of popular classifiers and emphasizes the value of cross-validation in medical prediction tasks. The proposed framework advances the field by identifying optimal models for real-world diabetes risk assessment.

Keywords:  K-Nearest Neighbors; Naive Bayes; cross validation; diabetes; diabetes mellitus; machine learning techniques; prediction; supervised

DOI:  https://doi.org/10.3389/fmed.2025.1620268
Front Endocrinol (Lausanne). 2025 ;16 1657366

Characterizing clinical risk profiles of major complications in type 2 diabetes mellitus using deep learning algorithms.

Haochen Liu, Xiaomiao Li, Ke Shi, Fengyu Lei, Ziyan Wang, Ziyuan Gao, Yunxi Liu, Jing Zhu, Jiajia Zhai, Yi Zhang, Xinyu Li, Shiyu Wang, Yu Niu, Louyan Ma, Tianxiao Zhang.

   Objective: To develop a self-reportable risk assessment tool for elderly type 2 diabetes mellitus (T2DM) patients, evaluating risks of diabetic nephropathy (DN), retinopathy (DR), peripheral neuropathy (DPN), and diabetic foot (DF) using machine learning, thereby providing new insights and tools for the screening and intervention of these complications.
Materials and methods: Data from 1,448 T2DM patients at Xi'an No.9 Hospital were used. After preprocessing, five machine learning algorithms (XGBoost, LightGBM, Random Forest, TabPFN, CatBoost) were applied. Models were trained on 70% of the data and evaluated on 30%, with performance assessed by multiple metrics and SHAP analysis for feature importance.
Results: The analysis identified 33 risk factors, including 6 shared risk factors (UACR for DN and DR; diabetes duration for DR, DPN, and DF; IBILI for DF and DPN; history of DN for DR and DF; U-Cr for DR and DF; MCHC for DN and DPN) and 27 unique risk factors. Model performance was robust: for DN, TabPFN achieved an AUC of 0.905 and Random Forest an accuracy of 0.878; for DR, LightGBM attained an AUC of 0.794; for DPN, both TabPFN and CatBoost achieved a perfect recall of 1.000 and F1-score of 0.915; and for DF, LightGBM attaining the highest AUC of 0.704. SHAP analysis highlighted key features for each complication, such as UACR and Y-protein for DN, diabetes duration and TPOAB for DR, history of DN and IBILI for DF, and diabetes duration and SBP for DPN.
Conclusion: This study employed interpretable machine learning to characterize risk factor profiles for multiple T2DM complications, identifying both common and distinct factors associated with major complications. The findings provide a foundation for exploring personalized risk management strategies and highlight the potential of data-driven approaches to inform early intervention research in T2DM complications.

Keywords:  SHAP (Shapley Additive explanation); Type 2 diabetes mellitus (T2DM); diabetic complications; machine learning; risk factors

DOI:  https://doi.org/10.3389/fendo.2025.1657366
Front Med (Lausanne). 2025 ;12 1644456

Artificial intelligence in proliferative diabetic retinopathy: advancing diagnosis, precision surgery, and anti-VEGF therapy optimization.

Geng-Qian Ke, Yang-Jun Fu, Zhuo-Han Huang, Shi-Xue Dai, Yun-Hua Wen, Hai-Xiang Lv.

  Proliferative diabetic retinopathy (PDR) represents the most advanced and vision-threatening stage of diabetic retinopathy (DR) and remains a leading cause of blindness in individuals with diabetes. This review presents a comprehensive overview of recent advances in the application of artificial intelligence (AI) for the diagnosis and treatment of PDR, emphasizing its clinical potential and associated challenges. The role of vascular endothelial growth factor (VEGF) in the pathogenesis of PDR has become increasingly clear, and AI offers novel capabilities in retinal image analysis, disease progression prediction, and treatment decision-making. These advancements have notably improved diagnostic accuracy and efficiency. Furthermore, AI-based models show promise in optimizing anti-VEGF therapy by enhancing therapeutic outcomes while reducing unnecessary healthcare expenditures. Future research should focus on the safe, effective, and ethical integration of AI into clinical workflows. Overcoming practical deployment barriers will require interdisciplinary collaboration among technology developers, clinicians, and regulatory bodies. The strategies and frameworks discussed in this review are expected to provide a foundation for future AI research and clinical translation in fields of PDR.

Keywords:  anti-VEGF therapy; artificial intelligence; deep learning; machine learning; proliferative diabetic retinopathy

DOI:  https://doi.org/10.3389/fmed.2025.1644456
BMJ Open. 2025 Sep 25. 15(9): e099062

Ratio of haemorrhagic area to retinal area as a novel indicator for AI-based screening of diabetic retinopathy in type 2 diabetes: a community-based cross-sectional study.

Ronsong Sun, Tianlin Zhang, Su Zhao, Zongli Hu, Junhua Wang, Bo Jiang, Qiying Pan, Yi Yang, Yuangdong Hu.

   BACKGROUND: The application of artificial intelligence (AI) technology in the screening of diabetic retinopathy (DR) has made significant strides. However, there remains a lack of comprehensive validation and evaluation of AI-derived quantitative indicators in DR screening.
OBJECTIVE: This study aims to assess the diagnostic performance of retinal microvascular indicators in the early detection of DR in patients with type 2 diabetes and to identify potential novel indicators for early DR screening.
RESEARCH DESIGN AND METHODS: This cross-sectional study included 533 community-recruited patients with type 2 diabetes mellitus who underwent fundus imaging. Based on the results of the fundus examination, the eyes were categorised into non-DR, mild non-proliferative diabetic retinopathy (NPDR), moderate NPDR and severe NPDR groups. AI systems were employed to quantify various retinal microvascular indicators, including microaneurysms (MAs), haemorrhage count (HC), haemorrhagic area (HA), the ratio of HA to retinal area (HA/RA), the ratio of HA to MA (HA/MA) and HC and/or MA (H/MA). Multivariable logistic regression was used to analyse the association between fundus indicators and DR severity, and receiver operating characteristic (ROC) curve analysis was performed to assess the predictive and screening value of these indicators, determining sensitivity, specificity, ROC area under the curve (AUC) and optimal cut-off values.
RESULTS: Among the 533 participants (mean age 64.03±9.71 years; 51.6% female), the DR prevalence was 10.0%. After adjusting for age, gender, body mass index, hypertension, diabetes duration, glycated haemoglobin levels, smoking and alcohol consumption, multivariable logistic regression indicated that HA/RA (OR 1.873, 95% CI 1.453 to 2.416) and HA/MA (OR 1.115, 95% CI 1.063 to 1.169) were associated with mild NPDR. Similarly, HA/RA (OR 1.928, 95% CI 1.509 to 2.464) and HA/MA (OR 1.165, 95% CI 1.112 to 1.220) were associated with moderate NPDR, and HA/RA (OR 2.435, 95% CI 1.921 to 3.086) and HA/MA (OR 1.171, 95% CI 1.117 to 1.226) were linked to severe NPDR. ROC curve analysis revealed that before adjustment, HA/RA demonstrated the highest screening value for DR, with an AUC of 0.917, sensitivity of 86.14%, specificity of 93.41%, Youden's index of 0.796 and an optimal cut-off value of 0.063. After adjusting for confounding factors, the AUC for HA/RA in diagnosing DR was 0.900, with sensitivity of 83.17%, specificity of 86.28%, Youden's index of 0.695 and an optimal cut-off value of 0.093.
CONCLUSIONS: The HA/RA and HA/MA show robust screening performance for early DR. These indicators should be considered for inclusion in AI-based early DR screening systems in the future.

Keywords:  Artificial Intelligence; Diabetes Mellitus, Type 2; Diabetic retinopathy; Observational Study

DOI:  https://doi.org/10.1136/bmjopen-2025-099062
PLoS One. 2025 ;20(9): e0332442

Diabetic retinopathy as the primary predictor of mild cognitive impairment in type 2 diabetes: Insights from machine learning models.

Fatima Zahra Rhmari Tlemçani, Faiçal Aitlahbib, Soukaina Laidi, Najib Alidrissi, Jehanne Aasfara, Saloua Elamari, Asma Chadli, Imane Motaib.

Mild cognitive impairment (MCI) is a significant and increasingly recognized problem in individuals with type 2 diabetes mellitus (T2DM). This study aims to develop a machine-learning model to predict MCI in patients with T2DMThe dataset was obtained from a prospective cohort conducted at the Sheikh Khalifa Ibn Zaid Hospital, Casablanca, Morocco, and was randomly split into three parts. Machine learning models were trained using the training dataset, and their performance was assessed on the test dataset. The unseen data was reserved for final validation. Subsequently, recursive feature elimination was applied with the top two algorithms to identify and retain the most impactful features for predicting MCI. Then, we retrained the models using the selected variables. Finally, the variables most contributing to the prediction of MCI were represented in a Shapley Additive Explanations SHAPE value plot to better understand their contribution and their ranking in MCI prediction.The dataset included 100 patients. Extra Trees classifier was the best-performing model for mild cognitive impairment (Accuracy: 0·9310/AUC: 0·9667/ Recall: 0·9333/ Precision: 0·9333/F1: 0·9333). The most contributing factors to MCI in patients with type 2 diabetes were, respectively, diabetic retinopathy, age, serum LDL cholesterol level, microalbuminuria, HbA1c, and, serum creatinine level. Our findings suggest avenues for early intervention that could prevent the progression of cognitive impairment among patients with type 2 diabetes.

DOI: https://doi.org/10.1371/journal.pone.0332442
PLoS One. 2025 ;20(9): e0332845

Retraction: Improved support vector machine based on CNN-SVD for vision-threatening diabetic retinopathy detection and classification.

PLOS One Editors

DOI: https://doi.org/10.1371/journal.pone.0332845
Life (Basel). 2025 Sep 08. pii: 1411. [Epub ahead of print]15(9):

HIRD-Net: An Explainable CNN-Based Framework with Attention Mechanism for Diabetic Retinopathy Diagnosis Using CLAHE-D-DoG Enhanced Fundus Images.

Muhammad Hassaan Ashraf, Muhammad Nabeel Mehmood, Musharif Ahmed, Dildar Hussain, Jawad Khan, Younhyun Jung, Mohammed Zakariah, Deema Mohammed AlSekait.

  Diabetic Retinopathy (DR) is a leading cause of vision impairment globally, underscoring the need for accurate and early diagnosis to prevent disease progression. Although fundus imaging serves as a cornerstone of Computer-Aided Diagnosis (CAD) systems, several challenges persist, including lesion scale variability, blurry morphological patterns, inter-class imbalance, limited labeled datasets, and computational inefficiencies. To address these issues, this study proposes an end-to-end diagnostic framework that integrates an enhanced preprocessing pipeline with a novel deep learning architecture, Hierarchical-Inception-Residual-Dense Network (HIRD-Net). The preprocessing stage combines Contrast Limited Adaptive Histogram Equalization (CLAHE) with Dilated Difference of Gaussian (D-DoG) filtering to improve image contrast and highlight fine-grained retinal structures. HIRD-Net features a hierarchical feature fusion stem alongside multiscale, multilevel inception-residual-dense blocks for robust representation learning. The Squeeze-and-Excitation Channel Attention (SECA) is introduced before each Global Average Pooling (GAP) layer to refine the Feature Maps (FMs). It further incorporates four GAP layers for multi-scale semantic aggregation, employs the Hard-Swish activation to enhance gradient flow, and utilizes the Focal Loss function to mitigate class imbalance issues. Experimental results on the IDRiD-APTOS2019, DDR, and EyePACS datasets demonstrate that the proposed framework achieves 93.46%, 82.45% and 79.94% overall classification accuracy using only 4.8 million parameters, highlighting its strong generalization capability and computational efficiency. Furthermore, to ensure transparent predictions, an Explainable AI (XAI) approach known as Gradient-weighted Class Activation Mapping (Grad-CAM) is employed to visualize HIRD-Net's decision-making process.

Keywords:  Computer-Aided Diagnosis; Convolutional Neural Network; Difference of Gaussian; Grad-CAM; Hard-Swish; deep learning in medical imaging; diabetic retinopathy diagnosis; explainable AI; fundus image enhancement

DOI:  https://doi.org/10.3390/life15091411
Life (Basel). 2025 Sep 19. pii: 1473. [Epub ahead of print]15(9):

Interpreting Deep Neural Networks in Diabetic Retinopathy Grading: A Comparison with Human Decision Criteria.

Sangeeta Biswas, Md Ahanaf Arif Khan, Md Hasnain Ali, Johan Rohdin, Subrata Pramanik, Md Iqbal Aziz Khan, Sanjoy Kumar Chakravarty, Bimal Kumar Pramanik.

  Diabetic retinopathy (DR) causes visual impairment and blindness in millions of diabetic patients globally. Fundus image-based Automatic Diabetic Retinopathy Classifiers (ADRCs) can ensure regular retina checkups for many diabetic patients and reduce the burden on the limited number of retina experts by referring only those patients who require their attention. Over the last decade, numerous deep neural network-based algorithms have been proposed for ADRCs to distinguish the severity levels of DR. However, it has not been investigated whether DNN-based ADRCs consider the same criteria as human retina professionals (HRPs), i.e., whether they follow the same grading scale when making decisions about the severity level of DR, which may put the reliability of ADRCs into question. In this study, we investigated this issue by experimenting on publicly available datasets using MobileNet-based ADRCs and analyzing the output of the ADRCs using two eXplainable artificial intelligence (XAI) techniques named Gradient-weighted Class Activation Map (Grad-CAM) and Integrated Gradients (IG).

Keywords:  Grad-CAM; d iabetic retinopathy classification; deep neural network; fundus image; integrated gradients

DOI:  https://doi.org/10.3390/life15091473
Diabetes Obes Metab. 2025 Sep 22.

GlucoNet-MM: A multimodal attention-based multi-task learning framework with decision transformer for personalised and explainable blood glucose forecasting.

Sarmad Maqsood, Muhammad Abdullah Sarwar, Egle Belousovienė, Rytis Maskeliūnas.

   AIMS: Accurate and personalized blood glucose prediction is critical for proactive diabetes management. Conventional machine learning (ML) models often struggle to generalize across patients due to individual variability, nonlinear glycemic dynamics, and sparse multimodal input data. This study aims to develop an advanced, interpretable deep learning (DL) framework for patient-specific, policy-aware blood glucose forecasting.
MATERIALS AND METHODS: We propose GlucoNet-MM, a novel multimodal DL framework that combines attention-based multi-task learning (MTL) with a Decision Transformer (DT), a reinforcement learning paradigm that frames policy learning as sequence modeling. The model integrates heterogeneous physiological and behavioral data, continuous glucose monitoring (CGM), insulin dosage, carbohydrate intake, and physical activity, to capture complex temporal dependencies. The MTL backbone learns shared representations across multiple prediction horizons, while the DT module conditions future glucose predictions on desired glycemic outcomes. Temporal attention visualizations and integrated gradient-based attribution methods are used to provide interpretability, and Monte Carlo dropout is employed for uncertainty quantification.
RESULTS: GlucoNet-MM was evaluated on two publicly available datasets, BrisT1D and OhioT1DM. The model achieved R2 scores of 0.94 and 0.96 and mean absolute error (MAE) values of 0.031 and 0.027, respectively. These results outperform single-modality and conventional non-adaptive baseline models, demonstrating superior predictive accuracy and generalizability.
CONCLUSION: GlucoNet-MM represents a promising step toward intelligent, personalized clinical decision support for diabetes care. Its multimodal design, policy-aware forecasting, and interpretability features enhance both prediction accuracy and clinical trust, enabling proactive glycemic management tailored to individual patient needs.

Keywords:  blood glucose forecasting; decision transformer; explainable artificial intelligence; multimodal deep learning; multi‐task learning

DOI:  https://doi.org/10.1111/dom.70147
JMIR Diabetes. 2025 Sep 25. 10 e69142

Toward a Clinically Actionable, Electronic Health Record-Based Machine Learning Model to Forecast 90-Day Change in Hemoglobin A1c in Youth With Type 1 Diabetes: Feasibility and Model Development Study.

Erin M Tallon, David D Williams, Cintya Schweisberger, Colin Mullaney, Brent Lockee, Diana Ferro, Craig A Vandervelden, Mitchell S Barnes, Angelica Cristello Sarteau, Anna R Kahkoska, Susana R Patton, Sanjeev Mehta, Ryan McDonough, Marcus Lind, Leonard D'Avolio, Mark A Clements.

   Background: Clinicians currently lack an effective means for identifying youth with type 1 diabetes (T1D) who are at risk for experiencing glycemic deterioration between diabetes clinic visits. As a result, their ability to identify youth who may optimally benefit from targeted interventions designed to address rising glycemic levels is limited. Although electronic health records (EHR)-based risk predictions have been used to forecast health outcomes in T1D, no study has investigated the potential for using EHR data to identify youth with T1D who will experience a clinically significant rise in glycated hemoglobin (HbA1c) ≥0.3% (approximately 3 mmol/mol) between diabetes clinic visits.
Objective: We aimed to evaluate the feasibility of using routinely collected EHR data to develop a machine learning model to predict 90-day unit-change in HbA1c (in % units) in youth (aged 9-18 y) with T1D. We assessed our model's ability to augment clinical decision-making by identifying a percent change cut point that optimized identification of youth who would experience a clinically significant rise in HbA1c.
Methods: From a cohort of 2757 youth with T1D who received care from a network of pediatric diabetes clinics in the Midwestern United States (January 2012-August 2017), we identified 1743 youth with 9643 HbA1c observation windows (ie, 2 HbA1c measurements separated by 70-110 d, approximating the 90-day time interval between routine diabetes clinic visits). We used up to 5 years of youths' longitudinal EHR data to transform 17,466 features (demographics, laboratory results, vital signs, anthropometric measures, medications, diagnosis codes, procedure codes, and free-text data) for model training. We performed 3-fold cross-validation to train random forest regression models to predict 90-day unit-change in HbA1c(%).
Results: Across all 3 folds of our cross-validation model, the average root-mean-square error was 0.88 (95% CI 0.85-0.90). Predicted HbA1c(%) strongly correlated with true HbA1c(%) (r=0.79; 95% CI 0.78-0.80). The top 10 features impacting model predictions included postal code, various metrics related to HbA1c, and the frequency of a diagnosis code indicating difficulty with treatment engagement. At a clinically significant percent rise threshold of ≥0.3% (approximately 3 mmol/mol), our model's positive predictive value was 60.3%, indicating a 1.5-fold enrichment (relative to the observed frequency that youth experienced this outcome [3928/9643, 40.7%]). Model sensitivity and positive predictive value improved when thresholds for clinical significance included smaller changes in HbA1c, whereas specificity and negative predictive value improved when thresholds required larger changes in HbA1c.
Conclusions: Routinely collected EHR data can be used to create an ML model for predicting unit-change in HbA1c between diabetes clinic visits among youth with T1D. Future work will focus on optimizing model performance and validating the model in additional cohorts and in other diabetes clinics.

Keywords:  AI, artificial intelligence; EHR, electronic health records; HbA1c, hemoglobin A1c; T1D, type 1 diabetes; adolescent; clinical decision support; glycemic control; machine learning; pediatric; population health; prediction; real-world data; youth

DOI:  https://doi.org/10.2196/69142
Vision (Basel). 2025 Sep 01. pii: 75. [Epub ahead of print]9(3):

A Comparative Study Between Clinical Optical Coherence Tomography (OCT) Analysis and Artificial Intelligence-Based Quantitative Evaluation in the Diagnosis of Diabetic Macular Edema.

Camila Brandão Fantozzi, Letícia Margaria Peres, Jogi Suda Neto, Cinara Cássia Brandão, Rodrigo Capobianco Guido, Rubens Camargo Siqueira.

  Recent advances in artificial intelligence (AI) have transformed ophthalmic diagnostics, particularly for retinal diseases. In this prospective, non-randomized study, we evaluated the performance of an AI-based software system against conventional clinical assessment-both quantitative and qualitative-of optical coherence tomography (OCT) images for diagnosing diabetic macular edema (DME). A total of 700 OCT exams were analyzed across 26 features, including demographic data (age, sex), eye laterality, visual acuity, and 21 quantitative OCT parameters (Macula Map A X-Y). We tested two classification scenarios: binary (DME presence vs. absence) and multiclass (six distinct DME phenotypes). To streamline feature selection, we applied paraconsistent feature engineering (PFE), isolating the most diagnostically relevant variables. We then compared the diagnostic accuracies of logistic regression, support vector machines (SVM), K-nearest neighbors (KNN), and decision tree models. In the binary classification using all features, SVM and KNN achieved 92% accuracy, while logistic regression reached 91%. When restricted to the four PFE-selected features, accuracy modestly declined to 84% for both logistic regression and SVM. These findings underscore the potential of AI-and particularly PFE-as an efficient, accurate aid for DME screening and diagnosis.

Keywords:  artificial intelligence; diabetic macular edema; machine learning; optical coherence tomography; paraconsistent feature engineering; retinal diseases; support vector machine

DOI:  https://doi.org/10.3390/vision9030075
Genes (Basel). 2025 Sep 16. pii: 1096. [Epub ahead of print]16(9):

Precision Medicine for Diabetic Retinopathy: Integrating Genetics, Biomarkers, Lifestyle, and AI.

Connor Kaurich, Neha Mahajan, Ashay D Bhatwadekar.

  Diabetic retinopathy (DR) is a common sight-threatening complication of diabetes. Overall, 26% of the 37 million Americans with diabetes suffer from DR, and 5% of people with diabetes suffer from vision-threatening DR. DR is a heterogeneous disease; thus, it is essential to acknowledge this diversity as we advance toward precision medicine. The current classification for DR primarily focuses on the microvascular component of disease progression, which does not fully capture the heterogeneity of disease etiology in different patients. Further, researchers in the field have suggested renewed interest in improving diagnosis and treatment modalities for personalized care in DR management. Moreover, genetic factors, lifestyle, and environmental variation strongly affect the disease outcome. It is important to emphasize that various ocular and peripheral biomarkers, along with imaging techniques, significantly influence the diagnosis of DR. Therefore, in this review, we explore the heterogeneity of DR, genetic variations or polymorphism, lifestyle and environmental factors, and how these factors may affect the development of precision medicine for DR. First, we provide a rationale for the necessity of a multifaceted understanding of disease etiology. Next, we discuss different aspects of DR diagnosis, emphasizing the need for further stratification of patient populations to facilitate personalized treatment. We then discuss different genetics, race, sex, lifestyle, and environmental factors that could help personalize treatments for DR. Lastly, we summarize the available literature to elaborate how artificial intelligence can enhance diagnostics and disease classification and create personalized treatments, ultimately improving disease outcomes in DR patients who do not respond to first-line care.

Keywords:  diabetic retinopathy; personalized medicine; precision medicine

DOI:  https://doi.org/10.3390/genes16091096
J Med Internet Res. 2025 Sep 24. 27 e69408

Diabetic Foot Ulcer Classification Models Using Artificial Intelligence and Machine Learning Techniques: Systematic Review.

Manuel Alberto Silva, Emma J Hamilton, David A Russell, Fran Game, Sheila C Wang, Sofia Baptista, Matilde Monteiro-Soares.

   BACKGROUND: Diabetes-related foot ulceration (DFU) is a common complication of diabetes, with a significant impact on survival, health care costs, and health-related quality of life. The prognosis of DFU varies widely among individuals. The International Working Group on the Diabetic Foot recently updated their guidelines on how to classify ulcers using "classical" classification and scoring systems. No system was recommended for individual prognostication, and the group considered that more detail in ulcer characterization was needed and that machine learning (ML)-based models may be the solution. Despite advances in the field, no assessment of available evidence was done.
OBJECTIVE: This study aimed to identify and collect available evidence assessing the ability of ML-based models to predict clinical outcomes in people with DFU.
METHODS: We searched the MEDLINE database (PubMed), Scopus, Web of Science, and IEEE Xplore for papers published up to July 2023. Studies were eligible if they were anterograde analytical studies that examined the prognostic abilities of ML models in predicting clinical outcomes in a population that included at least 80% of adults with DFU. The literature was screened independently by 2 investigators (MMS and DAR or EH in the first phase, and MMS and MAS in the second phase) for eligibility criteria and data extracted. The risk of bias was evaluated using the Quality In Prognosis Studies tool and the Prediction model Risk Of Bias Assessment Tool by 2 investigators (MMS and MAS) independently. A narrative synthesis was conducted.
RESULTS: We retrieved a total of 2412 references after removing duplicates, of which 167 were subjected to full-text screening. Two references were added from searching relevant studies' lists of references. A total of 11 studies, comprising 13 papers, were included focusing on 3 outcomes: wound healing, lower extremity amputation, and mortality. Overall, 55 predictive models were created using mostly clinical characteristics, random forest as the developing method, and area under the receiver operating characteristic curve (AUROC) as a discrimination accuracy measure. AUROC varied from 0.56 to 0.94, with the majority of the models reporting an AUROC equal or superior to 0.8 but lacking 95% CIs. All studies were found to have a high risk of bias, mainly due to a lack of uniform variable definitions, outcome definitions and follow-up periods, insufficient sample sizes, and inadequate handling of missing data.
CONCLUSIONS: We identified several ML-based models predicting clinical outcomes with good discriminatory ability in people with DFU. Due to the focus on development and internal validation of the models, the proposal of several models in each study without selecting the "best one," and the use of nonexplainable techniques, the use of this type of model is clearly impaired. Future studies externally validating explainable models are needed so that ML models can become a reality in DFU care.
TRIAL REGISTRATION: PROSPERO CRD42022308248; https://www.crd.york.ac.uk/PROSPERO/view/CRD42022308248.

Keywords:  artificial intelligence; classification; diabetic foot; machine learning; prognosis.

DOI:  https://doi.org/10.2196/69408
Prim Care Diabetes. 2025 Sep 25. pii: S1751-9918(25)00191-3. [Epub ahead of print]

Predicting the future risk of developing type 2 diabetes in women with a history of gestational diabetes mellitus using machine learning and explainable artificial intelligence.

Jenifar Prashanthan, Amirthanathan Prashanthan.

   BACKGROUND AND AIM: It is essential to identify the risk of developing Type 2 Diabetes Mellitus (T2DM) in women with a history of Gestational Diabetes Mellitus (GDM). This study seeks to create a machine learning (ML) model combined with explainable artificial intelligence (XAI) to predict and explain the risk of Type 2 Diabetes Mellitus (T2DM) in women with a history of Gestational Diabetes Mellitus (GDM).
METHODS: A literature review found 28 risk factors, including pregnancy-related clinical risk factors, maternal characteristics, genetic risk factors, and lifestyle and modifiable risk factors. A synthetic dataset was generated utilizing subject expertise and clinical experience through Python programming. Various machine learning classification techniques were employed on the data to identify the optimal model, which integrates interpretability approaches (SHAP) to guarantee the transparency of model predictions.
RESULTS: The developed machine learning model exhibited superior accuracy in predicting the risk of T2DM relative to conventional clinical risk scores, with notable contributions from factors such as insulin treatment during pregnancy, physical inactivity, obesity, breastfeeding, a history of recurrent GDM, an unhealthy diet, and ethnicity. Integrated XAI assists clinicians in comprehending the relevant risk factors and their influence on certain predictive outcomes.
CONCLUSIONS: Machine learning and explainable artificial intelligence provide a comprehensive methodology for individualized risk evaluation in women with a history of gestational diabetes mellitus. This methodology, by integrating extensive real-world data, offers healthcare clinicians actionable insights for early intervention.

Keywords:  Explainable AI; Gestational diabetes mellitus; Machine learning; Personalized healthcare; Risk prediction; Type 2 diabetes mellitus

DOI:  https://doi.org/10.1016/j.pcd.2025.09.006
Front Med (Lausanne). 2025 ;12 1610114

Patient acceptance of AI-assisted diabetic retinopathy screening in primary care: findings from a questionnaire-based feasibility study.

Malene Krogh, Marie Germund Nielsen, Gabriela Byskov Petersen, Morten Sig Ager Jensen, Martin Bach Jensen, Henrik Vorum, Niels Henrik Bruun, Jette Kolding Kristensen.

   Aim: This feasibility study investigates patients' acceptance of AI-assisted diabetic retinopathy screening (DRS) in primary care.
Method: Patients with type 2 diabetes from 12 primary care settings in Denmark underwent AI-assisted DRS as part of routine diabetes care and completed a questionnaire covering demographics, recent DRS, general health, mental well being, trust in physician, competence in diabetes self-care, distrust in AI, and acceptance of future DRS.
Results: 298 patients participated and completed the questionnaire. Acceptance of future AI-assisted DRS in primary care was higher than that of ophthalmologist-led screening, although patients still showed distrust toward AI. Findings indicated that greater competence in diabetes self-care was associated with higher acceptance of future AI-assisted DRS in primary care. Lower distrust in AI increased acceptance of future AI-assisted DRS in primary care, while higher distrust increased acceptance of ophthalmologist-led DRS.
Conclusion: This study found that most patients accepted future AI-assisted DRS in primary care. Associations between acceptance and the factors examined are very small and may have limited or no clinical impact. Other factors, such as convenience of having DRS in primary care, may influence patient's acceptance.

Keywords:  artificial intelligence; diabetic retinopathy screening; patient acceptance; primary care; questionnaire development; type 2 diabetes

DOI:  https://doi.org/10.3389/fmed.2025.1610114
Bioengineering (Basel). 2025 Aug 25. pii: 914. [Epub ahead of print]12(9):

Comprehensive Survey of OCT-Based Disorders Diagnosis: From Feature Extraction Methods to Robust Security Frameworks.

Alex Liew, Sos Agaian.

  Optical coherence tomography (OCT) is a leading imaging technique for diagnosing retinal disorders such as age-related macular degeneration and diabetic retinopathy. Its ability to detect structural changes, especially in the optic nerve head, has made it vital for early diagnosis and monitoring. This paper surveys techniques for ocular disease prediction using OCT, focusing on both hand-crafted and deep learning-based feature extractors. While the field has seen rapid growth, a detailed comparative analysis of these methods has been lacking. We address this by reviewing research from the past 20 years, evaluating methods based on accuracy, sensitivity, specificity, and computational cost. Key diseases examined include glaucoma, diabetic retinopathy, cataracts, amblyopia, and macular degeneration. We also assess public OCT datasets widely used in model development. A unique contribution of this paper is the exploration of adversarial attacks targeting OCT-based diagnostic systems and the vulnerabilities of different feature extraction techniques. We propose a practical, robust defense strategy that integrates with existing models and outperforms current solutions. Our findings emphasize the value of combining classical and deep learning methods with strong defenses to enhance the security and reliability of OCT-based diagnostics, and we offer guidance for future research and clinical integration.

Keywords:  adversarial attacks; clinical decision support systems; deep learning models; diabetic retinopathy; glaucoma detection; hand-crafted features; optical coherence tomography (OCT); robustness in medical imaging; security in AI model

DOI:  https://doi.org/10.3390/bioengineering12090914
Medicine (Baltimore). 2025 Sep 19. 104(38): e44815

Determining the association of C-reactive-protein-triglyceride-glucose index and diabetes using machine learning and LASSO regression: A cross-sectional analysis of NHANES 2001 to 2010 results.

Lin Xi, Jijing Zhao, Yunpeng Wang.

  The C-reactive protein-triglyceride-glucose index (CTI) has emerged as a novel metric for evaluating the severity of inflammation and the degree of insulin resistance. Nevertheless, the precise correlation between CTI and diabetes remains to be elucidated. Consequently, in this study, we elucidate the relationship between CTI and diabetes. The study utilized data from the National Health and Nutrition Examination Survey spanning from 2001 to 2010. To evaluate the association between CTI and the risk of diabetes, the research employed weighted logistic regression, subgroup analyses, and restricted cubic spline. Subsequently, participants were randomly assigned to the training and validation cohorts in a 7:3 ratio. Least Absolute Shrinkage and Selection Operator (LASSO) regression was employed to evaluate the validation cohort, select the optimal model, and identify potential confounding factors. The variables identified by LASSO regression were used to construct a nomogram-based predictive model, receiver operating characteristic curve, calibration curve, and decision curve analysis curve. The variables selected by LASSO regression were also incorporated into the ML model, and SHAP visualization analysis was performed. Upon adjustment for potential confounding factors, a significant positive correlation was observed between the CTI and the incidence of diabetes (OR = 1.96, 95% CI: 1.69-2.26, P < .001). Restricted cubic spline showed a linear positive correlation between CTI and incidence of diabetes mellitus (P-nonlinear = .5200). A total of 8 variables were identified through LASSO regression, including age, race, marital status, hypertension, body mass index, cardiovascular disease (CVD), and CTI. A nomogram-based predictive model was constructed using these predictors. The area under the receiver operating characteristic curve (AUC) in the validation cohort was 0.92, indicating a robust performance of the model. These 8 variables were subsequently incorporated into the ML model, and the CatBoost model demonstrated the best performance with an AUC of 0.843 (95% CI: 0.820-0.866). SHAP analysis revealed that hypertension was the most influential factor. A significant positive linear correlation was observed between higher CTI values and increased diabetes risk, suggesting that CTI has the potential to serve as a predictor for the incidence risk of diabetes.

Keywords:  CTI; LASSO regression; NHANES; SHAP; diabetes; machine learning; nomogram

DOI:  https://doi.org/10.1097/MD.0000000000044815
Front Med (Lausanne). 2025 ;12 1646495

AI-based integration of ECG biomarkers for assessing cardiac risk in type 2 diabetes mellitus with comorbid conditions for patient stratification.

Symeon Savvopoulos, Haralampos Hatzikirou, Herbert F Jelinek.

   Introduction: The increasing prevalence of type 2 diabetes mellitus (T2DM) requires improved early detection strategies that integrate demographic, clinical, physiological, and pharmacological data. Electrocardiographic (ECG) biomarkers offer a non-invasive means to assess diabetes-related cardiac risk, particularly in individuals with hypertension (HT) and cardiovascular disease (CVD) comorbidities of diabetes.
Methods: ECG data from 581 subjects were categorized by glycemic status (healthy, prediabetes, T2DM) and comorbidities. Demographic, clinical, and pharmaceutical data were merged with 10 s and 5 min ECG recordings. SMOTE was used to correct class imbalance. Support Vector Machines (SVM) performed best among machine learning classifiers. Classification accuracy, sensitivity, specificity, and AUC were computed using 5-fold cross-validation. Feature importance was assessed through permutation analysis to identify the most discriminative ECG and medication-related predictors.
Results: T2DM patients, particularly those with HT and CVD, exhibited significant prolongation of QTc (10 s), QTd (10 s and 5 min), and PQ intervals, as well as changes in the QRS-Axis, indicating increased arrhythmic risk and electrical remodeling (p < 0.001). Antihypertensive and lipid-lowering medications influenced QRS-Axis and PQ intervals, while antidepressant use was associated with QTd dispersion (p = 0.010). Classification accuracy ranged from 0.64 to 0.88. Five-minute ECGs provided higher accuracy (~0.88) when medication data were included, while 10-s ECGs performed well in treated patients (~0.86-0.88).
Discussion: This study shows that ECG-based, AI-driven screening captures the interaction between comorbidities, medication use, and cardiac electrophysiology. Integrating ECG biomarkers with medication data improved T2DM risk classification, enabling better treatment outcomes based on clinical use of non-invasive methods for risk classification.

Keywords:  arrhythmia medications; cardiovascular risk; diabetes; electrocardiography (ECG); hypertension disease progression; screening

DOI:  https://doi.org/10.3389/fmed.2025.1646495
Diabetes Metab Syndr Obes. 2025 ;18 3539-3552

A Scoping Review of Artificial Intelligence-Based Health Education Interventions for Patients with Type 2 Diabetes.

Chen Li, Wei Li, Yanru Shao, Zhigang Xu, Junyan Song, Yan Wang.

   Background: Type 2 diabetes mellitus (T2DM) poses a critical global health burden, requiring effective health education to enhance patient self-management. Artificial intelligence (AI) offers personalized and scalable solutions; however, comprehensive syntheses of its applications in T2DM health education are scarce.
Objective: Guided by the Arksey and O'Malley scoping review framework, this study maps AI-based health education interventions for T2DM by evaluating technologies, effectiveness, and challenges.
Methods: Seven academic databases (PubMed, Web of Science, Embase, Scopus, EBSCO, the Cochrane Library, the Joanna Briggs Institute (JBI) Database, and Wiley Online Library) were searched for studies published from 2008 to March 2025, identifying 14 eligible interventional studies involving 32,478 adult T2DM patients receiving AI-based health education.
Results: (1) Technological Diversity: Interventions included mobile apps (eg, FoodLens, TRIO system), chatbots, intelligent platforms, and machine learning algorithms, focusing on diet, glucose monitoring, and lifestyle management. (2) Effectiveness: AI interventions enhanced glycemic control, yielding reductions in glycosylated hemoglobin (HbA1c) of up to 2.59%, improved self‑management adherence (60-85%), and produced positive psychological outcomes (eg, increased self‑efficacy); efficacy varied by intervention duration and user engagement. (3) Challenges: Key barriers included technical complexity, low long-term engagement, digital literacy gaps, and data privacy concerns.
Conclusion: AI holds substantial potential for T2DM health education via personalized, accessible interventions. Future research should address technological hurdles, prioritize user-centered design, and integrate AI into healthcare systems to ensure sustainability and equity.

Keywords:  AI; T2DM; health education; scoping review

DOI:  https://doi.org/10.2147/DMSO.S541515
J Neurol Sci. 2025 Sep 15. pii: S0022-510X(25)00321-1. [Epub ahead of print]478 123701

Multimodal MRI analysis selecting key brain features for machine learning based classification of diabetic neuropathic pain and phenotypes.

Søren Nf Hostrup, Henrik P Lind, Suganthiya S Croosu, Johan Røikjer, Carsten D Mørch, Niels Ejskjær, Lasse R Østergaard, Tine M Hansen, Jens B Frøkjær.

  Cerebral alterations are associated with diabetic peripheral neuropathy (DPN) and neuropathic pain, including reductions in brain volumes, cortical thickness, sulcus depth, and alterations in metabolites and functional connectivity. This study combined multimodal magnetic resonance imaging (MRI) data to differentiate clinical phenotypes and uncover distinct associated brain features using machine-learning based classification. Seventy-six participants were recruited: 20 healthy and 56 with type 1 diabetes mellitus: 18 without DPN, 19 with painless DPN, and 19 with painful DPN. Three machine learning classifiers were evaluated, including accompanying feature importance from the highest performing classifier. Class membership probabilities (predicted probability of belonging to a group) were compared across classes and correlated with clinical measures of pain and nerve function for the class concerned. Accuracies were ≥ 0.75 for all classes except painless DPN, though separated by painful membership probability (p ≤ 0.02). Overall classification accuracy was 0.71. The most informative features were functional connectivity, followed by N-acetylaspartate/creatine and sulcal depth. The painful DPN group was separated from the remaining groups by painful membership probability (p ≤ 0.01), which correlated with pain measures. Diabetes without DPN membership probability correlated with sural nerve conduction (p ≤ 0.001, rs≥0.49) and warm detection thresholds (p ≤ 0.001, rs= - 0.51). This exploratory study suggests that different MRI modalities provide complementary information describing phenotypes of diabetes, DPN and DPN related pain, with functional connectivity being most essential. Thus, implying a multifactorial cerebral manifestation of neuropathy and pain in diabetes and aiding the development of grading and prognostic tools for personalised treatments. Studies of larger cohorts should validate these findings.

Keywords:  Brain signatures; Diabetic neuropathy; Machine learning classification; Magnetic resonance imaging; Neuropathic pain; Type 1 diabetes

DOI:  https://doi.org/10.1016/j.jns.2025.123701