bims-aukdir 2026-01-11 papers

bims-aukdir

Biomed News

on Automated knowledge discovery in diabetes research

Issue of 2026–01–11
twenty-one papers selected by
Mott Given

Hybrid quantum-classical deep learning framework for balanced multiclass diabetic retinopathy classification.
Mapping neighbourhood-level drivers of type 2 diabetes for precision public health using predictive and causal machine learning.
Machine Learning Analysis of Retrospective Data From 503 Hospitalized Older Patients With Type 2 Diabetes to Identify Factors Associated With Cognitive Impairment.
Assessment of Blood Glucose Measurement Using New Noninvasive Technology: Protocol and Methodology.
DHAN: A deep hierarchical attention network for multisource biomedical data fusion in precision diabetic retinopathy diagnosis.
Development and validation of a multi-modality system combining radiomics and deep learning for predicting mid-pregnancy complications and enabling timely pregnancy care.
Diabetic foot ulcer classification using an enhanced coordinate attention integrated ConvNext model.
Quantitative evaluation of retinal vascular morphology based on the human visual bionic mechanism for the evaluation of diabetic retinopathy onset.
Integrated biomarker analysis and next-generation AI for precision diabetes prediction.
Predicting Chronic Kidney Disease in Type 2 Diabetes Using Natural Language Processing on Healthcare Data.
An integrated design for segmentation and classification of diabetic foot ulcers using thermography images.
Machine Learning Identification and Animal Model Validation of Key Genes for Lipid Metabolism in Diabetic Nephropathy.
Integrative Multi-Omics and Machine Learning Reveal Shared Biomarkers in Type 2 Diabetes and Atherosclerosis.
Non-Glycemic Clinical Data for Type 2 Diabetes Detection in Mexican Adults: A Comparative Analysis of Atherogenic Indices, Statistical Transformations, and Machine Learning Algorithms.
GraphRAG-Enabled Local Large Language Model for Gestational Diabetes Mellitus: Development of a Proof-of-Concept.
Development of a risk assessment model for multimorbidity of diabetes, hypertension, and coronary heart disease with XGBoost in primary care in Shanghai, China: a retrospective study.
Cyber-Physical-Human Systems in Precision Medicine: Advances in Artificial Pancreas for Treatment of Diabetes.
Technology-driven diabetes management: a review of digital therapeutics, mobile apps, and clinical evidence.
Artificial Intelligence in Diabetic Kidney Disease Research: Bibliometric Analysis From 2006 to 2024.
The effects of e-Health and artificial intelligence literacy levels on disease self-management in patients with diabetes.
Discovery of PPARγ Partial Agonists for Treatment of Type 2 Diabetes Based on an Integrated Virtual Screening Strategy that Combines Fragment Molecular Orbital Calculations, Machine Learning, Molecular Docking, Interaction Fingerprint Filtering, and Molecular Dynamics Simulations.

MethodsX. 2025 Dec;15 103605

Hybrid quantum-classical deep learning framework for balanced multiclass diabetic retinopathy classification.

Tabassum Ara, Ved Prakash Mishra, Manish Bali, Anuradha Yenkikar.

  Diabetic Retinopathy (DR) is a progressive eye disease and a leading cause of preventable blindness among diabetic patients. Early and accurate classification of its severity stages is crucial for effective treatment but remains challenging due to class imbalance, high-resolution data, and limited scalability of existing models. This study presents a novel hybrid quantum-classical deep learning framework to address these limitations in five-class DR classification. The model achieves a balanced accuracy of 80.96 % on the APTOS 2019 dataset, outperforming several classical baselines across all DR stages. It is optimized for computational efficiency and class-balanced learning, making it suitable for deployment in telemedicine platforms and low-resource clinical settings. This work contributes a scalable AI-based diagnostic approach that fuses deep learning with emerging quantum computing techniques. The methodology, results, and publicly shared codebase provide a replicable framework for researchers and practitioners working in AI for medical imaging and early disease screening. This method is well-suited for low-resource clinical environments and tele-ophthalmology applications. The method involves an:•ResNet-50 feature extractor with a 4-stage dense projection (2048→8) for quantum-ready compression•8-qubit VQC with parameterized RY-RZ gates and ring-style entanglement for high expressiveness•Stratified sampling + mixed-precision training for efficiency and class-balanced generalization.

Keywords:  Diabetic retinopathy detection; Multiclass medical image classification; Quantum machine learning, Hybrid quantum-classical model; ResNet50

DOI:  https://doi.org/10.1016/j.mex.2025.103605
Sci Rep. 2026 Jan 05.

Mapping neighbourhood-level drivers of type 2 diabetes for precision public health using predictive and causal machine learning.

Mohammad Noaeen, Amirhosein Rostami, Ibrahim Ghanem, Olli Saarela, Karim Keshavjee, Jeffrey R Brook, Zahra Shakeri.

Type 2 diabetes has become an urban epidemic influenced by neighbourhood environments. However, conventional risk models focusing solely on individual factors fail to account for these neighbourhood influences and often require detailed patient data that may not be available. To address this gap, we developed an integrated approach combining machine learning and causal inference to map type 2 diabetes risk at the neighbourhood level. Using demographic, health, and socioeconomic data from 1,149 Census Tracts (CTs; the neighbourhood unit in this study) in a large metropolitan region, we trained seven machine learning models to identify neighbourhoods with high diabetes prevalence. Although neighbourhood-level diabetes data were available for this study area, our model's high predictive accuracy on external validation data (area under the curve (AUC) = 0.95), particularly from a distinct geographical region, suggests potential utility for predicting diabetes risk in other Canadian regions or elsewhere where such data are unavailable, provided comparable covariates are available and the model is locally retrained and validated using spatially aware procedures. The top models achieved high recall ([Formula: see text]) and AUC up to 0.96 on test data, indicating accurate identification of high-risk neighbourhoods with few missed high-risk areas. Survey-derived neighbourhood health indicators, including obesity rate, physical inactivity, and median age were strong predictors of diabetes prevalence. We then applied a Causal Forest approach to estimate conditional average treatment effects (CATE, τ) for selected potentially modifiable factors and summarized the results with the mean [Formula: see text]. Higher work stress ([Formula: see text]) and daily smoking ([Formula: see text]) were moderately associated with increased risk, whereas better mental health ([Formula: see text]) was protective, highlighting mental health as a priority for further evaluation, especially in neighbourhoods predicted to have high diabetes prevalence. These findings could help identify modifiable neighbourhood-level factors for local prevention efforts and inform equity-oriented planning in diverse urban populations. Prospective or quasi-experimental studies are needed to evaluate intervention effects. Our integrated machine-learning and causal framework lays the groundwork for precision public health, suggesting that modifiable neighbourhood factors may indicate diabetes risk when patient-level data are scarce. Furthermore, the pipeline is conceptually adaptable to other chronic diseases influenced by social and environmental determinants and may inform targeted prevention beyond type 2 diabetes, contingent on disease-specific feature sets and external validation.

DOI: https://doi.org/10.1038/s41598-025-34287-7
Med Sci Monit. 2026 Jan 10. 32 e949864

Machine Learning Analysis of Retrospective Data From 503 Hospitalized Older Patients With Type 2 Diabetes to Identify Factors Associated With Cognitive Impairment.

Mingzhu Yu, Jianfeng Zhang, Haigeng Chen, Guiyue Li.

BACKGROUND Diabetes is increasingly prevalent among older adults; mild cognitive impairment (MCI) comorbidity in this group represents a major concern. Existing MCI prediction methods are often inaccurate, but machine learning (ML) offers improved potential. This study aimed to identify factors associated with MCI through ML analysis of retrospective data from hospitalized older patients with type 2 diabetes mellitus (T2DM). MATERIAL AND METHODS This retrospective study analyzed data from 503 inpatients older than 60 years with T2DM. Patients were classified into MCI (n=102) and normal (n=401) groups based on Mini-Mental State Examination scores. To minimize overfitting and maximize data utilization, 5-fold cross-validation was used for model training and evaluation. Least absolute shrinkage and selection operator regression identified 8 core predictors from clinical data. Logistic regression, eXtreme Gradient Boosting (XGBoost), and random forest algorithms were employed to construct predictive models. Receiver operating characteristic (ROC) curves were used to compare model performance. RESULTS Key predictors of early MCI included age, body mass index, glycated hemoglobin, C-reactive protein, waist-to-height ratio, presence of diabetic complications, diabetes duration exceeding 5 years, and low education level. The XGBoost model outperformed other algorithms in ROC analysis: area under the curve, 0.892±0.032; accuracy, 0.851±0.028; sensitivity, 0.843±0.031; specificity, 0.859±0.029; and F1 score, 0.834±0.033. CONCLUSIONS The XGBoost model, incorporating these identified factors, demonstrated optimal predictive performance for MCI in older patients with T2DM. It may aid clinical risk stratification and provide a quantitative foundation for early intervention.

DOI: https://doi.org/10.12659/MSM.949864
JMIR Res Protoc. 2026 Jan 08. 15 e76558

Assessment of Blood Glucose Measurement Using New Noninvasive Technology: Protocol and Methodology.

Eka Widrian Suradji, Satvinder Singh Dhaliwal, Zhang Li-Feng, Entong Zhou, Amos Lim, Hui Hern Wong, Yao Du, Fredicia Fredicia, Amanda Dianky, Puspasari Pasaribu, Ernest Gamaliel Masayoshi Takashi, Justin Yong An Tan, Mardi Santoso, Seng Bin Ang.

   Background: Diabetes mellitus (DM) is a major noncommunicable disease with a significant increase in prevalence, especially in low- and middle-income countries. The latest International Diabetes Federation Diabetes Atlas (2025) reports that 11.1% of the adult population (20 to 79 years old) is living with diabetes, with over 4 in 10 unaware of their condition. Early diagnosis and treatment of diabetes reduce the risk and slow the progression of debilitating complications, such as amputation, vision loss, renal failure, cardiovascular disease, dementia, some cancers, and infections like tuberculosis and severe COVID-19. Current screening methods for diabetes are invasive and costly. This has limited their utilization, especially in high-density populations and low- and middle-income countries such as Indonesia. Blood Glucose Evaluation and Monitoring (BGEM) is a machine learning algorithm developed by Actxa to analyze photoplethysmography data from wearable devices for diabetic risk assessment. Its noninvasive and user-friendly nature makes it a strong candidate for fulfilling the need for a diabetes screening or monitoring tool.
Objective: The aim of this study is to collect a large and more diverse dataset for the training of BGEM machine learning models. This dataset is intended to improve the model's generalizability and to evaluate its performance across different age groups, racial groups, and skin types, with the goal of enhancing accuracy and robustness for diabetes risk assessment and glucose monitoring.
Methods: Adult participants aged 18 years and above, with either a diabetic or a nondiabetic history, who reside in Greater Jakarta Area, Indonesia, were approached for recruitment. Blood glucose was assessed using laboratory blood analysis from capillary or plasma samples after fasting and at 1, 2, and 3 hours after a meal. BGEM data were also collected at each of these time points. Anthropological measurements with a standardized questionnaire on physical activity, demographic information, respondent's diabetic status, and current medications taken were also collected.
Results: Between June and October 2024, 885 participants were enrolled. Eight photoplethysmography recordings per participant were collected across 4 meal time points using 2 wearable devices in addition to the collection of clinical measurements, blood sampling, and related questionnaires. .
Conclusions: This protocol paper outlines the methodology designed for assessing and interpreting participants' blood sugar profiles, especially on demographic variability, in order to evaluate BGEM, a photoplethysmography-based artificial intelligence model designed to estimate blood glucose levels and diabetic risk. The clinical trial was conducted on Indonesian participants with and without diabetes while considering various influencing factors. This dataset is designed to enable assessment of the model's performance across diverse racial, risk factors, and skin-type groups, with the aim of making the model more valid and reliable.

Keywords:  BGEM; Blood Glucose Evaluation and Monitoring; Indonesia; artificial intelligence; blood glucose; diabetes mellitus; machine learning; noninvasive; photoplethysmography; screening; wearables

DOI:  https://doi.org/10.2196/76558
Digit Health. 2026 Jan-Dec;12:12 20552076251408517

DHAN: A deep hierarchical attention network for multisource biomedical data fusion in precision diabetic retinopathy diagnosis.

Rui Tao, Hongru Li, Jingyi Lu, Yaxin Wang, Jian Zhou, Xia Yu.

   Objectives: To establish a robust and clinically applicable approach for integrating heterogeneous multisource biomedical data, particularly continuous glucose monitoring (CGM) profiles and structured electronic health records (EHRs), in order to enhance the diagnostic accuracy and clinical utility of diabetic retinopathy (DR) detection.
Methods: This study proposed a deep hierarchical attention network (DHAN) for multisource biomedical data fusion. First, to address the heterogeneous forms of different data sources, two specific subencoders were designed, a hybrid architecture for time-series CGM sensors and a structured encoder for EHRs. Second, an entity-embedding mechanism was added to the EHR subencoder to fuse heterogeneous feature types within EHRs. Finally, a deep hierarchical attention mechanism was proposed to dynamically capture inner-source saliency and inter-source correlations.
Results: Using the dataset provided by Shanghai Sixth People's Hospital, 559 patients were included, comprising 157 with DR and 402 without. DHAN achieved the best performance across multiple experiments, with a diagnostic accuracy of 0.89. Its comprehensive performance, including an F1-score of 0.80 and a G-mean of 0.89, further demonstrates its robustness.
Conclusions: The results indicate that DHAN is a viable approach for diagnosing DR in patients with type 2 diabetes. By effectively fusing multisource heterogeneous data, DHAN can be embedded within CGM sensors to enable remote concurrent diagnosis of DR. Moreover, it provides a generalizable paradigm for multisensor systems requiring fusion of data from multiple sources.

Keywords:  Continuous glucose monitoring sensors; hierarchical attention; medical signal processing; multisource biomedical data; personalized medicine

DOI:  https://doi.org/10.1177/20552076251408517
Front Pediatr. 2025 ;13 1716073

Development and validation of a multi-modality system combining radiomics and deep learning for predicting mid-pregnancy complications and enabling timely pregnancy care.

Juan Guo, Yuhong Huang, Zhiwei Zhang, Baoqiang Shi, Shuxian Xi, Yuanyuan Mai, Yan Liang, Zhizhen Guo, Lantian Shang.

   Introduction: To improve the early prediction of hypertensive disorders of pregnancy (HDP) and gestational diabetes mellitus (GDM), we developed and validated an artificial intelligence (AI) model. This initiative was driven by the insufficient accuracy of current clinical tools. Our study aimed to determine whether integrating radiomics and deep learning features from first-trimester ultrasound scans could enhance predictive performance.
Methods: A total of 213 pregnant women who underwent ultrasound at 8 weeks of gestation were enrolled. Clinical data, radiomics features, and deep learning features were collected. Imaging features were selected using LASSO regression. Four predictive models were developed: a clinical model, a radiomics model, a deep learning model, and a fusion model combining all feature types. Model performance was evaluated on an independent test set using metrics including AUC, sensitivity, specificity, calibration, and decision curve analysis.
Results: In the training cohort, all models demonstrated excellent discriminatory ability, with the combined model achieving the highest AUC of 0.987 (95% CI: 0.9733-0.9999), followed by the DLR model (AUC = 0.985). The clinical model (AUC = 0.941) and radiomics model (AUC = 0.939) also performed well. In the test cohort, the combined model maintained superior performance with an AUC of 0.963 (95% CI: 0.9152-1.0000), significantly outperforming all single-modality models. Overall, the combined model exhibited optimal and stable predictive performance across both training and test datasets.
Discussion: This enables accurate early prediction of HDP and GDM. This non-invasive tool supports tailored prenatal care, with potential to improve outcomes. Further validation in diverse groups is needed.

Keywords:  deep learning; gestational diabetes mellitus; hypertensive disorders of pregnancy; radiomics; ultrasound

DOI:  https://doi.org/10.3389/fped.2025.1716073
Phys Eng Sci Med. 2026 Jan 05.

Diabetic foot ulcer classification using an enhanced coordinate attention integrated ConvNext model.

L Jani Anbarasi, R Neeraja, S Geetha, R Vidhya, Vinayakumar Ravi, D Dhanya.

  Diabetic foot ulcers (DFUs) pose a significant complication of diabetes with the potential to lead to amputation if not effectively managed. Current DFU treatments require rigorous monitoring by both healthcare professionals and patients, which is challenging due to the high costs associated with diagnosis, treatment and long-term care. A major limitation of these approaches is their limited capacity to identify highly relevant pattern connections and broad contextual correlations resulting inaccuracies in classifying regions of interest. This research introduces an attention enhanced deep learning-based automated approach for assessing DFUs using images to expedite the investigation process and offer optimal recommendations. Adaptive thresholding is employed to enhance the contrast and uniformity of DFU images and thereby improves the feature extraction. A hybrid model incorporating coordinate attention enhanced ConvNeXt is used for effective DFU image classification to enhance the representation of complex patterns through efficient parameter utilization. The ConvNeXt architecture is designed to scale efficiently across various sizes by utilizing depthwise separable convolutions and improved image normalization. This model is augmented with coordinate attention, which captures spatial information in both horizontal and vertical directions, aiding in the extraction of long-range dependency features for more accurate classification of DFU images. Experimental results demonstrate that the model achieves an accuracy of 97.16% and F1-score of 0.97.

Keywords:  Adaptive thresholding; ConvNeXt model; Coordinate attention model; Diabetic foot ulcer

DOI:  https://doi.org/10.1007/s13246-025-01692-1
Microvasc Res. 2026 Jan 06. pii: S0026-2862(26)00001-4. [Epub ahead of print] 104901

Quantitative evaluation of retinal vascular morphology based on the human visual bionic mechanism for the evaluation of diabetic retinopathy onset.

Yi Xu, Yuting Wu, Saiguang Ling, Zhou Dong, Xin Ke, Lina Lu, Zheng Ye, Jianling Song, Haidong Zou.

   BACKGROUND AND AIMS: A limited amount of diabetic retinopathy (DR) development can be explained by traditional risk factors. This study aimed to determine the association of artificial intelligence (AI)-assisted retinal vasculature measurement parameters with DR onset in adults with type 2 diabetes.
METHODS: This observational cohort study was conducted in 556 patients with type 2 diabetes without DR who underwent general and ophthalmological examinations. Their blood pressure, body mass index (BMI), fasting blood glucose (FBG), and glycosylated hemoglobin levels were measured. An AI-based fundus image analysis system was used to assess vessel tortuosity, fractal dimension, and retinal arteriolar/venular diameters in different regions.
RESULTS: At the end of the observation period, 299 patients remained free of DR (control group), whereas 257 developed DR (progression group). The retinal arteriolar caliber, venular caliber, arteriolar tortuosity, and venular tortuosity did not differ significantly between the groups at baseline (P > 0.05). However, DR onset was significantly correlated with retinal arteriolar caliber, fractal dimensions, and retinal venular tortuosity (P < 0.05). The widening of the retinal arteriolar diameter within the 1.5-2.0 PD region of the optical disc center was the strongest predictor of DR development. It also improved the performance of the DR onset prediction model compared with those using traditional risk factors alone.
CONCLUSIONS: AI-assisted retinal vasculature measurements were associated with DR onset and progression. In addition to increased retinal venular tortuosity and fractal dimension, retinal arteriolar caliber within the 1.5-2.0 PD may serve as a valuable biomarker of early vascular dysfunction and increased DR risk.

Keywords:  Artificial intelligence; Diabetic retinopathy; Vessel measurement

DOI:  https://doi.org/10.1016/j.mvr.2026.104901
Toxicol Res. 2026 Jan;42(1): 35-46

Integrated biomarker analysis and next-generation AI for precision diabetes prediction.

Hyun Sim, Hyung-Ho Ha, Hangun Kim.

  This study explored the integration of advanced deep learning with key pharmaceutical biomarkers to enhance early diabetes prediction. We developed a multimodal ensemble approach that leverages transformer architectures to capture complex dependencies in heterogeneous healthcare data and Diffusion Models to address class imbalances by generating synthetic samples. Our research utilized diverse data sources, including electronic health records, medical imaging, and wearable device time-series data, supplemented with synthetic samples to better represent minority populations such as patients with type 1 and gestational diabetes. Critical biomarkers, including C-peptide, insulin, and hemoglobin A1c, were incorporated to improve model interpretability. The methodology involved extensive evaluation using accuracy, area under the receiver operating characteristic (ROC) curve (AUC), precision, recall, and F1-score, with cross-validation to mitigate overfitting. We also implemented interpretability features to provide clinicians with insight into the significance of biomarkers. Results showed a 6.2% improvement in minority class recall when pharmaceutical biomarkers were combined with diffusion-based augmentation. The model demonstrated enhanced classification stability and provided clear insights into clinical decision-making, highlighting the influence of biomarkers on disease progression and treatment outcomes. Future work will focus on multicenter validation, integration of additional omics data, and specialized validation across diverse populations. These findings underscore the potential of AI-driven biomarker analysis for advancing early diagnosis and personalized diabetes management, with broader implications for chronic disease prediction.

Keywords:  Biomarker analysis; Deep learning; Diabetes prediction; Diffusion model; Synthetic data augmentation

DOI:  https://doi.org/10.1007/s43188-025-00312-0
Kidney Dis (Basel). 2026 Jan-Dec;12(1):12(1): 18-28

Predicting Chronic Kidney Disease in Type 2 Diabetes Using Natural Language Processing on Healthcare Data.

Juan F Navarro-González, Leopoldo Pérez de Isla, Gloria Cánovas Molina, Miguel Ángel Brito-Sanfiel, David Emilio Barajas Galindo, Luís Ángel Cuellar Olmedo, Dídac Mauricio, Santiago Tofé Povedano, José Antonio Balsa Barro, Matilde Rubio Almanza, José Juan Aparicio Sánchez, Miren Sequera Mutiozabal, Belén Pimentel, Ana Pérez Domínguez, Carlos Arias-Cabrales, Víctor Fanjul, Antonio Jesús Blanco-Carrasco, Juan Francisco Merino Torres.

   Introduction: Persons with type 2 diabetes mellitus (T2DM) attending hospitals frequently experience major complications. We assessed the potential use of unstructured free-text data extracted from electronic health records (EHRs) using natural language processing (NLP) and machine learning (ML) to develop a predictive model for chronic kidney disease (CKD) in T2DM.
Methods: This multicenter retrospective study included data from eight Spanish hospitals (2013-2018), extracted using NLP and ML techniques (EHRead®) based on SNOMED CT terminology. From a cohort of individuals with T2DM, we identified those with and without CKD at inclusion. Among individuals without CKD, we trained and validated a 2-year predictive model for CKD development. The model showing the best balance between performance and clinical interpretability was selected for integration into a web-based tool to support early detection and risk stratification.
Results: Of 588,786 individuals with T2DM, 316,597 were included for model development (training: 291,429 [92.1%]; validation: 25,168 [7.9%]; CKD incidence: 15.4% and 18.4%, respectively). A high proportion of missing data was observed in key clinical variables. Among models evaluated, logistic regression achieved the best performance (receiver operating characteristic area under the curve 0.72) using 27 predictors. Both a reduced 10-predictor model and a clinically refined 8-predictor model showed comparable performance to the full model in training and validation cohorts. The clinically refined model was selected for implementation in the web-based tool.
Conclusion: Unstructured EHR data enabled the development of a predictive model for 2-year CKD risk in persons with T2DM. Improving EHR data completeness remains essential to enhance future predictive modeling.

Keywords:  Chronic kidney disease; Electronic health records; Machine learning; Natural language processing; Predictive model; Real-world data; Type 2 diabetes mellitus

DOI:  https://doi.org/10.1159/000547604
J Diabetes Metab Disord. 2026 Jun;25(1): 17

An integrated design for segmentation and classification of diabetic foot ulcers using thermography images.

Wan Suhaimizan Bin Wan Zaki, Yamunarani Thanikachalam, Ashok Vajravelu, Prabu Rathinam, E Thangavelu.

  In recent days, diabetics, a chronic disease has risen significantly which leads to more health complications. Among those complications, diabetic foot ulcer (DFU) is much serious. DFU is a wound on the foot of a person who is affected with diabetics. It sometimes leads to fatality if untreated. Diagnosing the DFU in its early stage remains challenging due to medical impediments by the diabetics. Thermography serves as a promising technique in the early prediction of the DFU and aids for an improvised treatment towards the eradication of foot amputations. But still, utilizing thermography images for clinical treatments continues to be underexplored in treating DFU due to its computational complexities and existence of ambiguities in thermal images. To overcome this challenge, this research paper proposes an Intelligent Prediction System (IPS) using the modified swin transformers for an effective segmentation and deep capsule networks for an accurate prediction of DFU. In the segmentation phase, swin transformers can be used as U-NET based architecture to segment the lesions of foot ulcers. Deep features are extracted by the capsule networks and supplied to the deep shallow network which works on the standard of extreme learning networks to achieve the early prediction of DFU. The extensive experimentation is conducted using the thermal foot ulcer images in Python3.20 and Tensorflow -Keras Libraries. To verify the efficiency of the proposed schema, evaluated performances are assessed with other research experiments. Results show that the proposed schema achieves the highest prediction accuracy (99%) with promising segmented performance (98.6%). Moreover, the proposed model excels the varied residing schema and establishes a firm foothold in the early prediction of DFUs.

Keywords:  Capsule networks; Diabetic foot ulcers; Extreme learning principles; Shallow network; Swin transformers

DOI:  https://doi.org/10.1007/s40200-025-01813-3
Zhongguo Yi Xue Ke Xue Yuan Xue Bao. 2025 Dec;47(6): 873-887

Machine Learning Identification and Animal Model Validation of Key Genes for Lipid Metabolism in Diabetic Nephropathy.

Yue Li, Yu-Yu Zhang, Xing Wan, Song-Lin Niu, Hong-Hong Shi, Li-Hua Wang.

  Objective To identify key genes of lipid metabolism in diabetic nephropathy(DN) through machine learning models and animal model validation. Methods The limma R package was used for differential gene expression analysis on 69 samples from two transcriptome datasets of the Gene Expression Omnibus and 2 184 differentially expressed genes were identified.Subsequently,we adopted undifferentiated consensus clustering to classify DN samples into two specific subtypes.At the same time,we performed weighted gene co-expression network analysis to mine the gene modules significantly associated with DN.In addition,using least absolute shrinkage and selection operator,support vector machine-recursive feature elimination,and random forest machine learning techniques,combined with protein-protein interaction network analysis,we screened out three core genes.Finally,we constructed a mouse model of type 2 diabetes mellitus to verify the effectiveness of the expression of these key genes. Results Three core genes,APOO,ALDH7A1,and ALB,were predicted as potential biomarkers of lipid metabolism in DN,and their expression levels were downregulated in DN.Through experimental validation in a diabetic mouse model,we confirmed the altered expression of APOO,ALDH7A1,and ALB in DN,which supported their potential as diagnostic markers. Conclusions Our findings suggest that APOO,ALDH7A1,and ALB are new diagnostic markers associated with lipid metabolism in DN,which provides new perspectives for understanding the molecular mechanisms of lipid metabolism in DN.

Keywords:  bioinformatics; biomarkers; diabetic nephropathy; lipid metabolism; machine learning

DOI:  https://doi.org/10.3881/j.issn.1000-503X.16594
Int J Mol Sci. 2025 Dec 22. pii: 136. [Epub ahead of print]27(1):

Integrative Multi-Omics and Machine Learning Reveal Shared Biomarkers in Type 2 Diabetes and Atherosclerosis.

Qingjie Wu, Zhaochu Wang, Mengzhen Fan, Linglun Hao, Jicheng Chen, Changwen Wu, Bizhen Gao.

  Atherosclerosis (AS) is a leading cause of death and disability in type 2 diabetes mellitus (T2DM). However, the shared molecular mechanisms linking T2DM and atherosclerosis have not been fully elucidated. We analyzed AS- and T2DM-related gene expression profiles from the Gene Expression Omnibus (GEO) database to identify overlapping differentially expressed genes and co-expression signatures. Functional enrichment (Gene Ontology (GO)/Kyoto Encyclopedia of Genes and Genomes (KEGG)) and protein-protein interaction (PPI) network analyses were then used to describe the pathways and interaction modules associated with these shared signatures, We next applied the cytoHubba algorithm together with several machine learning methods to prioritize hub genes and evaluate their diagnostic potential and combined CIBERSORT-based immune cell infiltration analysis with single-cell RNA sequencing data to examine cell types and the expression patterns of the shared genes in specific cell populations. We identified 72 shared feature genes. Functional enrichment analysis of these genes revealed significant enrichment of inflammatory- and metabolism-related pathways. Three genes-IL1B, MMP9, and P2RY13-emerged as shared hub genes and yielded robust ANN-based predictive performance across datasets. Immune deconvolution and single-cell analyses consistently indicated inflammatory amplification and an imbalance of macrophage polarization in both conditions. Biology mapped to the hubs suggests IL1B drives inflammatory signaling, MMP9 reflects extracellular-matrix remodeling, and P2RY13 implicates cholesterol transport. Collectively, these findings indicate that T2DM and AS converge on immune and inflammatory processes with macrophage dysregulation as a central axis; IL1B, MMP9, and P2RY13 represent potential biomarkers and therapeutic targets and may influence disease progression by regulating macrophage states, supporting translational application to diagnosis and treatment of T2DM-related atherosclerosis. These findings are preliminary. Further experimental and clinical studies are needed to confirm their validity, given the limitations of the present study.

Keywords:  atherosclerosis; immune infiltration; machine learning; single-cell; type 2 diabetes mellitus

DOI:  https://doi.org/10.3390/ijms27010136
Diagnostics (Basel). 2025 Dec 23. pii: 53. [Epub ahead of print]16(1):

Non-Glycemic Clinical Data for Type 2 Diabetes Detection in Mexican Adults: A Comparative Analysis of Atherogenic Indices, Statistical Transformations, and Machine Learning Algorithms.

Martin Hazael Guerrero-Flores, Valeria Maeda-Gutiérrez, Carlos E Galván-Tejada, Jorge I Galván-Tejada, Miguel Cruz, Luis Alberto Flores-Chaires, Karina Trejo-Vázquez, Rafael Magallanes-Quintanar, Javier Saldívar.

  Background: Type 2 diabetes (T2D) is a growing public health problem in Mexico. Lipid profile alterations have been shown to appear years before changes in glycemic biomarkers, and some of the latter are limited in availability, especially in underserved settings. Therefore, anthropometric variables and lipids represent relevant early indicators for the early detection of the disease. This study evaluates the capacity of non-glycemic clinical data-including lipid profile and anthropometric indicators-to detect T2D using machine learning, and compares the performance of different feature engineering approaches. Methods: Using more than a thousand clinical records of Mexican adults, three experiments were developed: (1) a distribution and normality analysis to characterize the variability of lipid variables; (2) an evaluation of the predictive power of multiple atherogenic indices (Castelli I, Castelli II, TG/HDL, and AIP); and (3) the implementation of statistical transformations (logarithmic, quare-root, and Z-standardization) to stabilize variance and improve feature quality. Logistic regression, SVM-RBF, random forest, and XGBoost models were trained on each feature set and evaluated using accuracy, sensitivity, specificity, F1-score, and area under the ROC curve. Results: The AIP index showed the greatest discriminatory power among the atherogenic indices, while normality-based transformations improved the performance of distribution-sensitive models, such as SVM. In the final experiment, the SVM-RBF and XGBoost models achieved AUC values greater than 0.90, demonstrating the feasibility of a diagnostic approach based exclusively on non-glycemic data. Conclusions: The findings indicate that the transformed lipid profile and anthropometric variables can constitute a solid and accessible alternative for the early detection of T2D in clinical and public health contexts, offering a robust methodological framework for future predictive applications in the absence of traditional glycemic biomarkers.

Keywords:  atherogenic indices; lipid profile; machine learning; non-glycemic biomarkers; type 2 diabetes

DOI:  https://doi.org/10.3390/diagnostics16010053
JMIR Diabetes. 2026 Jan 05. 11 e76454

GraphRAG-Enabled Local Large Language Model for Gestational Diabetes Mellitus: Development of a Proof-of-Concept.

Edmund Evangelista, Fathima Ruba, Salman Bukhari, Amril Nazir, Ravishankar Sharma.

   Background: Gestational diabetes mellitus (GDM) is a prevalent chronic condition that affects maternal and fetal health outcomes worldwide, increasingly in underserved populations. While generative artificial intelligence (AI) and large language models (LLMs) have shown promise in health care, their application in GDM management remains underexplored.
Objective: This study aimed to investigate whether retrieval-augmented generation techniques, when combined with knowledge graphs (KGs), could improve the contextual relevance and accuracy of AI-driven clinical decision support. For this, we developed and validated a graph-based retrieval-augmented generation (GraphRAG)-enabled local LLM as a clinical support tool for GDM management, assessing its performance against open-source LLM tools.
Methods: A prototype clinical AI assistant was developed using a GraphRAG constructed from 1212 peer-reviewed research articles on GDM interventions, retrieved from the Semantic Scholar API (2000-2024). The GraphRAG prototype integrated entity extraction, KG construction using Neo4j, and retrieval-augmented response generation. The performance was evaluated in a simulated environment using clinical and layperson prompts, comparing the outputs of the systems against ChatGPT (OpenAI), Claude (Anthropic), and BioMistral models across 5 common natural language generation metrics.
Results: The GraphRAG-enabled local LLM showed higher accuracy in generating clinically relevant responses. It achieved a bilingual evaluation understudy score of 0.99, Jaccard similarity of 0.98, and BERTScore of 0.98, outperforming the benchmark LLMs. The prototype also produced accurate, evidence-based recommendations for clinicians and patients, demonstrating its feasibility as a clinical support tool.
Conclusions: GraphRAG-enabled local LLMs show much potential for improving personalized GDM care by integrating domain-specific evidence and contextual retrieval. Our prototype proof-of-concept serves two purposes: (1) the local LLM architecture gives practitioners from underserved locations access to state-of-the-art medical research in the treatment of chronic conditions and (2) the KG schema may be feasibly built on peer-reviewed, indexed publications, devoid of hallucinations and contextualized with patient data. We conclude that advanced AI techniques such as KGs, retrieval-augmented generation, and local LLMs improve GDM management decisions and other similar conditions and advance equitable health care delivery in resource-constrained health care environments.

Keywords:  GDM; artificial intelligence; artificial intelligence for health care; explainable AI in medicine; generative AI; gestational diabetes mellitus; knowledge graph; large language model; retrieval augmented generation

DOI:  https://doi.org/10.2196/76454
BMC Cardiovasc Disord. 2026 Jan 05.

Development of a risk assessment model for multimorbidity of diabetes, hypertension, and coronary heart disease with XGBoost in primary care in Shanghai, China: a retrospective study.

Ning Chen, Peipei Zhang, Feng Fan, Hua Jin, Yanying Chen, Jing Liu, Lin Chen, Xiaoguang Wan, Jing Guo, Zaijia Yang, Lei Sun, Houqian Shan, Dehua Yu, Zhaoxin Wang, Jianwei Shi.

   BACKGROUND: Multimorbidity has emerged as a growing global health concern. Within its heterogeneous patterns, the cardiometabolic cluster is notably among the most common. Assessing the risk of such multimorbidity from a general practice perspective has become a priority in primary care. This study aimed to develop a comprehensive risk assessment model for the multimorbidity of diabetes, hypertension, and coronary heart disease among older adults in the community, utilizing large-scale data from Shanghai, China.
METHODS: Retrospective data spanning 2017 to 2019 were collected from 40,261 residents across 47 community health centers. These data comprised residents' health records, health examination results, hospital information system (HIS) records, imaging databases, and lifestyle information. The XGBoost machine learning algorithm was utilized to construct a comprehensive risk assessment model for the multimorbidity of diabetes, hypertension, and coronary heart disease. Model performance was evaluated using the area under the receiver operating characteristic curve, accuracy, precision, recall, and the F1 Score.
RESULTS: The dataset was split into training (80%) and testing (20%) sets. A total of 46 features were incorporated into the final comprehensive risk assessment model for the multimorbidity of diabetes, hypertension, and coronary heart disease. The optimal XGBoost model achieved a micro-average AUC of 0.822, a macro-average AUC of 0.795, and a weighted-average AUC of 0.784. These parameters demonstrate the high superiority of the constructed model.
CONCLUSIONS: The XGBoost-based risk assessment model for the multimorbidity of diabetes, hypertension, and coronary heart disease, integrated clinical and public health data from community residents. It identifies multidimensional predictors across four dimensions, underscoring its practical value in supporting integrated risk assessment and informing targeted health management strategies for individuals with multimorbidity.
CLINICAL TRIAL NUMBER: Not applicable.

Keywords:  Coronary heart disease; Diabetes; Hypertension; Multimorbidity; Primary care; Risk assessment model; XGBoost

DOI:  https://doi.org/10.1186/s12872-025-05479-w
Annu Rev Control. 2025 ;pii: 101033. [Epub ahead of print]60

Cyber-Physical-Human Systems in Precision Medicine: Advances in Artificial Pancreas for Treatment of Diabetes.

Mohammad Ahmadasas, Emirhan Inanc, Efe Ozkara, Mate Siket, Mudassir M Rashid, Mustafa Bilgic, Laurie Quinn, Ali Cinar.

  Cyber-physical-human systems (CPHS) hold the potential to transform healthcare delivery and patient outcomes for numerous chronic diseases, such as diabetes, through remote patient monitoring, automatic control, and precision medicine. Expedited by the recent advances in artificial intelligence, including model development, control systems, data assimilation, network infrastructure, and cybersecurity, the application of digital twins has proliferated across various industry sectors, and have recently been applied to medical settings. Digital twins of people with type 1 diabetes (T1D) and the pancreas can well represent the complex metabolic, physiologic, and pharmacologic processes underlying the chronic disease. This enables intelligent CPHSs that can automate insulin delivery without any manual user announcements to mitigate the effects of various disturbances to glucose homeostasis such as meals, physical activities, acute psychological stress, and sleep pattern variations. Automated insulin delivery in people with T1D, also called artificial pancreas, is a successful application of digital twins in medicine that advances T1D treatment, reducing the burden of the chronic condition and improving the lives of people with T1D. We present a hybrid modeling framework that integrates mechanistic physiological models with data-driven empirical models to develop accurate digital twins of people with T1D, which is then used in an artificial intelligence-enabled automated insulin delivery system. Simulation and clinical experiments integrating virtual patients or individuals with T1D with an artificial pancreas system illustrate the performance of the CPHS, demonstrating the capabilities of rendering fully-automated real-time treatment decisions for precision medicine. Future avenues of research and development in CPHS for precision medicine are also highlighted, including online learning algorithms, adaptive fault-tolerant systems, and robust cybersecurity.

Keywords:  Artificial pancreas; Cyber-physical-human systems; Data-driven model; Digital twin; Hybrid modeling; Model predictive control; Recursive least squares; Subspace identification; Type 1 diabetes

DOI:  https://doi.org/10.1016/j.arcontrol.2025.101033
Inform Health Soc Care. 2026 Jan 06. 1-21

Technology-driven diabetes management: a review of digital therapeutics, mobile apps, and clinical evidence.

Bidisha Bhattacharya, Kunal Kar, M P Venkatesh, Inderbir Singh.

  Digital health technologies are revolutionizing the treatment of diabetes by providing creative ways to enhance patient outcomes. In this study, the role of digital medicines in the treatment of diabetes is examined with special attention paid to wearable technology, mobile applications, and clinical data that backs up their usage. Software-driven interventions designed to prevent, manage, or treat diabetes using specialized data-driven methods are referred to as digital therapies. The article examines many smartphone applications that support insulin management, nutritional tracking, blood glucose monitoring, and lifestyle changes. It examines important clinical trials that demonstrate how successfully these tools reduce HbA1c levels, enhance glycemic control, and promote long-term treatment regimen adherence. New advancements in AI such as personalized AI algorithms, integration of continuous glucose monitors with mobile apps, remote patient monitoring, telemedicine, behavioral nudges, machine learning, and data analytics that enhance the personalization of diabetic care are also the subject of the evaluation. Digital treatments have the potential to revolutionize diabetes care by providing more accessible, patient-centered, and effective care.

Keywords:  AI and machine learning; Digital therapeutics; diabetes; mobile applications; wearable technology

DOI:  https://doi.org/10.1080/17538157.2025.2609747
JMIR Diabetes. 2026 Jan 09. 11 e72616

Artificial Intelligence in Diabetic Kidney Disease Research: Bibliometric Analysis From 2006 to 2024.

Xingyuan Li, Liming Xiao, Fenghao Yang, Fang Liu.

   Background: Diabetic kidney disease (DKD) is a major complication of diabetes and the leading cause of end-stage renal disease globally. Artificial intelligence (AI) technologies have shown increasing potential in DKD research for early detection, risk prediction, and disease management. However, the landscape of AI applications in this field remains incompletely mapped, especially in terms of collaboration networks, thematic evolution, and clinical translation.
Objective: This study aims to perform a comprehensive bibliometric and translational analysis of AI-related DKD research published between 2006 and 2024, identifying publication trends, research hotspots, key contributors, collaboration patterns, and the extent of clinical validation and explainability.
Methods: A systematic search of the Web of Science Core Collection was conducted to identify English-language original articles applying AI technologies to DKD. Articles were screened following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 guidelines. Bibliometric visualization was performed using CiteSpace and VOSviewer to assess coauthorship, institutional and country collaboration, keyword evolution, and citation bursts. A qualitative review was conducted to evaluate clinical validation, model explainability, and real-world implementation.
Results: Out of 1158 retrieved records, 384 studies met the inclusion criteria. Global publications on AI in DKD increased rapidly after 2019. China led in publication volume, followed by the United States, India, and Iran. Keyword analysis showed a thematic transition from early biomarker and proteomic research to deep learning, clinical prediction models, and management tools. Despite methodological advances, few studies included external validation or explainability frameworks. Notable translational efforts included DeepMind's acute kidney injury predictor and a chronic kidney disease prediction model developed by Sumit, yet widespread real-world integration remains limited.
Conclusions: AI research in DKD has grown substantially over the past 2 decades, with expanding international collaboration and diversification of research themes. However, challenges persist in clinical applicability, model transparency, and global inclusivity. Future research should prioritize explainable AI, multicenter validation, and integration into clinical workflows to support effective translation of AI innovations into DKD care.

Keywords:  artificial intelligence; bibliometric analysis; clinical validation; diabetic kidney disease; explainable AI; global collaboration

DOI:  https://doi.org/10.2196/72616
Diabetes Res Clin Pract. 2026 Jan 03. pii: S0168-8227(26)00001-X. [Epub ahead of print]232 113082

The effects of e-Health and artificial intelligence literacy levels on disease self-management in patients with diabetes.

Nilhan Töyer Şahin, Seda Pehlivan.

   BACKGROUND: This study was conducted to examine the effects of eHealth and artificial intelligence literacy on disease self-management in patients with diabetes.
METHODS: The cross-sectional study was conducted with 212 patients with diabetes who were followed up in Endocrinology clinics and outpatient clinics of a hospital between October 2024 and June 2025. Data were collected through face-to-face interviews using a Personal Information Form, the eHealth Literacy Scale, the Artificial Intelligence (AI) Literacy Scale, and the Diabetes Self-Management Questionnaire. Data were analysed using the SPSS-27 software, and p = 0.05 was considered statistically significant.
RESULTS: The mean age of the 212 patients was 52.09 ± 17.02, and their mean disease duration was 9.66 ± 8.47 years. The patients had mean Diabetes Self-Management Questionnaire, eHealth Literacy Scale, and AI Literacy Scale scores of 6.47 ± 1.50, 27.87 ± 8.83, and 48.12 ± 11.26, respectively.Diabetes self-management was significantly and positively correlated with eHealth literacy (r = 0.505; p = 0.000) and AI literacy (r = 0.499; p = 0.000). Additionally, a positive significant relationship was found between general eHealth literacy and AI literacy (r = 0.865; p = 0.000).
CONCLUSIONS: The results of this study suggest that general eHealth and AI literacy play a significant role in supporting diabetes self-management.

Keywords:  Artificial intelligence; Diabetes; Self-management; eHealth

DOI:  https://doi.org/10.1016/j.diabres.2026.113082
J Phys Chem B. 2026 Jan 06.

Discovery of PPARγ Partial Agonists for Treatment of Type 2 Diabetes Based on an Integrated Virtual Screening Strategy that Combines Fragment Molecular Orbital Calculations, Machine Learning, Molecular Docking, Interaction Fingerprint Filtering, and Molecular Dynamics Simulations.

Yulin Liu, Wei Liu, Risong Na, Xi Zhao.

Peroxisome proliferator-activated receptor γ (PPARγ) is a key therapeutic target for type 2 diabetes and cardiovascular diseases due to its central role in regulating glucose and lipid metabolism. While full PPARγ agonists exhibit efficacy, they are linked to adverse effects; in contrast, PPARγ partial agonists retain metabolic regulatory functions with improved safety, representing promising candidates for type 2 diabetes treatment. However, their action mechanisms and structure-activity relationships remain unclear. Herein, we developed an integrated virtual screening strategy combining fragment molecular orbital (FMO) calculations, machine learning, molecular docking, interaction fingerprint (IFP) filtering, and molecular dynamics (MD) simulations to identify potential PPARγ partial agonists and elucidate their interaction mechanisms. FMO analysis first confirmed interaction differences between PPARγ agonist classes at the binding pocket, pinpointing critical residues (CYS285, ARG288, ILE341, and SER342) for partial agonist activity. Using three machine learning algorithms (random forest, extra trees, and XGBoost) with extended connectivity fingerprints (ECFP), we constructed QSAR classification models and screened 9630 compounds. SHAP analysis highlighted key fingerprint fragments (positions 45, 1034, and 1243) governing bioactivity. Molecular docking and IFP refinement yielded six high-potency candidates, whose binding stability and partial agonist properties were validated via MD simulations, MM/PBSA binding free energy calculations, hydrogen bond analysis, and FMO calculations. Notably, these candidates did not directly interact with the AF2 domain, consistent with the canonical partial agonist mode of action. This multidisciplinary approach provides a framework for rational design of novel PPARγ partial agonists, and the identified molecules serve as promising leads for type 2 diabetes therapeutics.

DOI: https://doi.org/10.1021/acs.jpcb.5c06470