bims-aukdir 2026-04-05 papers

bims-aukdir

Biomed News

on Automated knowledge discovery in diabetes research

Issue of 2026–04–05
twenty-six papers selected by
Mott Given

Enhanced diagnosis of diabetic retinopathy: integrating advanced algorithms for automated detection and classification.
Diabetic retinopathy severity detection using an improved Whale optimization algorithm and convolutional Kolmogorov-Arnold network.
The transformative role of artificial intelligence in gestational diabetes mellitus: advancements in screening, management, and long-term outcome prediction.
Improving Diabetic Foot Care With Infrared Thermography and Artificial Intelligence: A Review.
Clinical applications of machine learning for infection assessment in diabetic foot ulcers.
Artificial intelligence for interpreting diabetes data using clinician-curated benchmarks.
Machine Learning-Based Prediction Model Construction for Type 2 Diabetes Mellitus: A Comparison of Algorithms and Multilevel Risk Factor Analysis.
A fundus image dataset for intelligent diabetic retinopathy system.
SoleFusion-Net: an explainable multimodal deep learning framework for diabetic foot syndrome classification in type II diabetes mellitus.
Interpretable machine learning in type-2 diabetes prediction in patients with depressive symptoms: Insights from mendelian randomization and physical activity.
OMICS data in the diagnosis of diabetic retinopathy: A comparison between transcriptome data and DNA methylation data.
Sequential pattern transformer (SPT): a generative and interpretable framework for predicting disease trajectories.
Artificial Intelligence in Diabetes Care.
Diabetic peripheral neuropathy identification using enface optical coherence tomography and multi-head attention deep learning algorithm.
FunduScope: a human-centered, machine learning-based interactive tool for training junior ophthalmologists in diabetic retinopathy detection.
Glucose forecasting and hypoglycemia forewarning in type 1 and type 2 diabetes using deep learning.
Early prediction of insulin requirement in gestational diabetes using a parsimonious LASSO model and a point-based risk score.
Machine Learning-Derived Predictive Risk Score for Prediabetes and Type 2 Diabetes Development in Parous Women.
Role of retinal biomarkers in diabetes detection and risk prediction: A systematic scoping review.
A systematic review of the use of artificial intelligence in mental health-based diabetes care: Current applications and future directions.
Evaluating ChatGPT for Carbohydrate Counting Accuracy in Diabetes Management: A Precision Health Approach.
Multimodal AI fuses proteomic and EHR data for rational prioritization of protein biomarkers in diabetic retinopathy.
Identification of mitochondria-related key genes in type 2 diabetes mellitus and elucidation of the Zhimu-Huangbai herb Pair's mechanism: an integrated approach of bioinformatics, machine learning, and experimental validation.
A Deep Learning Prognostic Model for Diabetes Patients Using Bilateral Fundus Imaging.
Integrating a Smart Sensor Chip and AI Predictive Analytics Into the Sehhaty App to Enhance Diabetes Management in Saudi Arabia.
Flexible statistical approaches for modeling nonlinear relationships in diabetes prediction using splines, Bayesian kernel regression and Bayesian regression trees.

Int J Ophthalmol. 2026 ;19(4): 637-645

Enhanced diagnosis of diabetic retinopathy: integrating advanced algorithms for automated detection and classification.

E Murali, T Prasad, J Hari Krishna, Abdul Hussain Sharief, Peta Nandini.

   AIM: To develop an automated diagnostic system for early detection of diabetic retinopathy (DR) using fundus images by identifying exudates, hemorrhages, and microaneurysms with advanced image processing and machine learning techniques.
METHODS: Fundus images from the IDRiD dataset and additional Kaggle datasets were used. A wavelet-based band-pass filter was applied for edge enhancement of retinal features. Gaussian mixture model (GMM) clustering was used to segment and extract texture features. These extracted features were classified using machine learning algorithms, including a random forest classifier and a multilayer perceptron neural network. Performance metrics such as sensitivity, specificity, and accuracy were computed to evaluate the proposed model's diagnostic effectiveness.
RESULTS: The random forest-based classification system achieved a sensitivity of 95.08%, specificity of 86.67%, and overall accuracy of 95.20% in detecting DR lesions. The combination of wavelet-based edge enhancement, GMM clustering, and neural network-based feature classification demonstrated high reliability in lesion identification.
CONCLUSION: The proposed method effectively detects early signs of DR from fundus images, offering a high-accuracy, automated, and scalable solution for assisting ophthalmologists. Its application can support large-scale screening programs, particularly in regions with limited access to specialized eye care.

Keywords:  Gaussian mixture modelclustering; diabetic retinopathy; fundus image analysis; multilayer perceptron; random forest algorithm

DOI:  https://doi.org/10.18240/ijo.2026.04.01
Front Med (Lausanne). 2026 ;13 1709872

Diabetic retinopathy severity detection using an improved Whale optimization algorithm and convolutional Kolmogorov-Arnold network.

Ashit Kumar Dutta, Nasser Ali Aljarallah, Abdul Rahaman Wahab Sait.

   Introduction: Diabetic retinopathy (DR) is an inflammatory condition affecting the retina caused by elevated and unregulated blood glucose levels. On a global scale, it is a contributing factor to vision impairment. Several deep learning (DL) methods use retinal images to identify DR severity. However, a significant improvement is required to assist medical professionals in recognizing DR in its early phases.
Methods: Thus, the author introduced a method based on the DL technique to determine the DR severity grades using retinal images. A ShuffleNet V2 model with vision transformers' (ViT) attention mechanism was used to extract the features. An improved Whale optimization method (IWO) was used to fine-tune the feature extraction model. We employed a convolutional Kolmogorov-Arnold Network to categorize the DR severity using the extracted features. The EyePACS dataset was utilized to train the proposed DR severity grading model using a five-fold cross-validation strategy. We generalized the model on the Messidor-2 dataset.
Results: The findings revealed an average accuracy of 93.84% on the MESSIDOR-2 dataset, demonstrating a substantial improvement in detecting DR using the fundus images.
Discussion: Furthermore, the model demands minimal processing resources to generate the outcomes, leading to the deployment of the proposed DR severity detection model in healthcare facilities with limited computational resources.

Keywords:  Kolmogorov-Arnold network; ShuffleNet V2; diabetic retinopathy; fundus images; hyperparameter optimization; improved Whale optimization; pre-trained models; transfer learning

DOI:  https://doi.org/10.3389/fmed.2026.1709872
J Diabetes Metab Disord. 2026 Jun;25(1): 120

The transformative role of artificial intelligence in gestational diabetes mellitus: advancements in screening, management, and long-term outcome prediction.

Zahra Taheri, Mobina Tajdari, Yekta Heidari Ghadikolaei, Hirad Eftekhari, Sarah Dodangeh, Porsa Jamshidi, Kamyab Andarzbakhsh, Amir Rezazadeh.

   Purpose: Artificial intelligence (AI) has emerged as a pivotal tool in enhancing the management of gestational diabetes mellitus (GDM). With its rising global prevalence, there is a pressing need for more precise, timely, and personalized strategies for diagnosis and treatment. This review highlights the role of AI in reshaping GDM management, providing actionable insights for researchers, clinicians, and policymakers.
Methods: Relevant literature was identified through a non-systematic search of databases including PubMed, Scopus, and Google Scholar. Articles published up to August 2025 were considered. Studies were selected based on their relevance to the topic. Key findings were organized thematically to highlight current trends, challenges, and future directions.
Results: The application of stacking-based machine learning models including random forest, AdaBoost, decision tree, support vector machine, naïve bayes, K-nearest neighbor, CatBoost, and XGBoost have facilitated the early detection of high-risk patients and achieved high precision and accuracy (88.8%) in reducing unnecessary diagnostic tests. Additionally, machine learning-based decision support systems have demonstrated promising capabilities in predicting insulin requirements and treatment failure with oral hypoglycemic drugs. The integration of AI in mobile applications and wearable devices has significantly encouraged proactive patient participation through real-time glucose monitoring and individualized feedback. Furthermore, these technologies are used to assess the long-term risk of developing type 2 diabetes postpartum, expanding the potential for innovative preventive strategies.
Conclusion: While challenges such as data heterogeneity, narrow applicability, and ethical concerns persist, there is an increasing body of evidence indicating that AI can play a revolutionary role in GDM management by providing personalized, cost-efficient, and scalable solutions.
Supplementary Information: The online version contains supplementary material available at 10.1007/s40200-025-01817-z.

Keywords:  Artificial intelligence; Digital health; Gestational diabetes mellitus; Machine learning; Personalized medicine; Predictive modeling

DOI:  https://doi.org/10.1007/s40200-025-01817-z
J Diabetes Sci Technol. 2026 Apr 03. 19322968261432639

Improving Diabetic Foot Care With Infrared Thermography and Artificial Intelligence: A Review.

Pedro Teixeira, Vítor Filipe, Ana Teixeira.

   BACKGROUND: One of the most common consequences in individuals with diabetes is the diabetic foot, which can cause foot ulcers and even lead to limb amputation. Since an increase of the temperature in the plantar region is directly correlated with an increased risk of ulceration, infrared thermography (IRT) has been used in multiple studies as an automatic tool for detecting problems in diabetic foot. Artificial intelligence-based computer-aided diagnosis systems are being more frequently used to improve decision-making and minimize errors. These technologies are designed to increase examination accuracy, consistency in image interpretation, prognosis evaluation support, and examination accuracy. They also have the ability to offer insightful information and help medical professionals to manage diabetic foot issues successfully.
METHODS: In this work, 37 papers that used thermography and artificial intelligence (AI) to identify diabetic foot complications and/or predict the risk of developing diabetic foot are analyzed.
RESULTS: The results demonstrate the potential of IRT imaging implementation with AI for the identification and prediction of diabetic foot complications.
CONCLUSIONS: The combination of IRT and AI shows significant potential for diabetic foot assessment; however, the great majority of these studies show that the research is confined to classification of foot thermograms using pre-prepared data sets. In particular, there is limited research on segmentation methods and constraints in the use of deep learning due to the lack of large and diverse datasets.

Keywords:  deep learning; diabetic foot; machine learning; neural network; thermography

DOI:  https://doi.org/10.1177/19322968261432639
Front Physiol. 2026 ;17 1776883

Clinical applications of machine learning for infection assessment in diabetic foot ulcers.

Khrystyna Lysnychka, Larysa Rostoka, Yana Burmistrova, Iryna Halabitska, Pavlo Petakh, Oleksandr Kamyshnyi.

  Diabetic foot ulcers (DFUs) represent one of the most severe complications of diabetes mellitus and are frequently complicated by infection, which significantly increases the risk of hospitalization, lower-limb amputation, and mortality. Early and accurate detection of infection in DFUs is therefore critical; however, clinical assessment remains challenging and is largely based on subjective visual evaluation. Inter-observer variability, atypical inflammatory responses in patients with diabetes, and inconsistent wound documentation contribute to delayed or inaccurate diagnoses. In recent years, digital sound imaging combined with machine learning (ML) techniques has emerged as a promising adjunct to traditional clinical assessment. This review summarizes and critically evaluates recent advances in the application of ML for infection assessment in DFUs. We review image-based ML approaches designed to detect infection-related visual features, including erythema, purulent exudate, necrosis, and tissue discoloration, as well as models developed for ulcer classification, tissue segmentation, and longitudinal wound monitoring. In addition, we discuss the clinical utility of ML-assisted tools in telemedicine, remote monitoring, and decision support, particularly in community and resource-limited settings. Current limitations, including image variability, dataset bias, lack of standardized imaging protocols, and limited clinical validation, are also addressed. Overall, ML-based systems have demonstrated encouraging performance in identifying infection-associated patterns in DFU images and may help reduce diagnostic variability and support earlier clinical intervention. Nevertheless, further large-scale prospective studies, regulatory validation, and integration into clinical workflows are required before widespread adoption. Machine learning should be viewed as a supportive tool that complements, rather than replaces, clinical expertise in the management of diabetic foot infections.

Keywords:  artificial intelligence; clinical decision support; diabetic foot infection; diabetic foot ulcer; digital wound imaging; machine learning; medical image analysis; telemedicine

DOI:  https://doi.org/10.3389/fphys.2026.1776883
Lancet Diabetes Endocrinol. 2026 Mar 27. pii: S2213-8587(26)00044-6. [Epub ahead of print]

Artificial intelligence for interpreting diabetes data using clinician-curated benchmarks.

David C Klonoff, Elizabeth Healey, Kenneth D Mandl, David Kerr.

DOI: https://doi.org/10.1016/S2213-8587(26)00044-6
J Diabetes Res. 2026 ;2026(1): e4525736

Machine Learning-Based Prediction Model Construction for Type 2 Diabetes Mellitus: A Comparison of Algorithms and Multilevel Risk Factor Analysis.

Qian Xu, Ruicong Yu, Huixin Qiu, Yannan Jiang, Jake Ball, Cuirong Xu, Jing Sun.

   BACKGROUND: Against the backdrop of the global high incidence of Type 2 diabetes mellitus (T2DM), existing prediction models are largely confined to single-dimensional risk factors, suffering from a core limitation of lacking multilevel integrated analysis. Given the severe impact of T2DM on individual health and healthcare systems, the construction of a comprehensive and accurate prediction model is of great significance.
OBJECTIVE: This study is aimed at constructing a T2DM prediction model, identifying multilevel risk factors, and enabling early screening, so as to help clinicians identify high-risk individuals and provide targets for public health interventions.
METHODS: Data from the National Health and Nutrition Examination Survey (NHANES) 2021-2023 were used, including 6337 participants aged 18 years and older. Missing values were handled using Monte Carlo multiple imputation, collinearity was reduced via principal component analysis (PCA), and feature selection was performed using random forest (RF) and recursive feature elimination (RFE). The adaptive synthetic sampling (ADASYN) method was applied to address class imbalance. The performance of seven machine learning models, including decision tree, random forest, extreme gradient boosting (XGBoost), and adaptive boosting (AdaBoost), was compared.
RESULTS: The AdaBoost model exhibited the optimal performance, with an area under the curve (AUC) of 0.85 (95% confidence interval: 0.85-0.86), an accuracy of 0.71 (95% confidence interval: 0.70-0.72), and an F1 score of 0.71; its performance was further improved after parameter optimization. A total of 24 key risk factors were identified, including 19 at the individual trait level, 3 at the individual behavior level, and 2 related to working and living conditions.
CONCLUSIONS: Machine learning models integrating multidimensional risk factors based on the health ecology framework can more accurately predict T2DM risk, providing a scientific basis for multilevel interventions. The innovation of this study lies in the first integration of the health ecology model with machine learning technology to systematically identify cross-level risk factors. Compared with traditional models, it is more comprehensive, breaks through the limitations of previous studies, and provides a new and effective tool for the precise prevention of T2DM and public health interventions.

Keywords:  Type 2 diabetes mellitus; health ecology; machine learning; prediction model

DOI:  https://doi.org/10.1155/jdr/4525736
Sci Data. 2026 Apr 01.

A fundus image dataset for intelligent diabetic retinopathy system.

Shaojuan Peng, Shuo Yang, Xinyu Zhao, Yongtao Zhang, Qingjie Bai, Duo Yuan, Yaling Liu, Yarou Hu, Yi Chen, Kaixuan Cui, Zhen Yu, Zhenquan Wu, Ruyin Tian, Baiying Lei, Guoming Zhang.

Diabetic retinopathy (DR), the most prevalent microvascular complication of diabetes mellitus, is the leading cause of irreversible vision loss in the global working-age population. At present, deep learning-integrated ultra-wide-field (UWF) image analysis systems have improved DR grading consistency and reduced peripheral lesion misdiagnosis rates, thereby overcoming the limitations of traditional 45° viewing field AI models in detecting peripheral retinal lesions. However, the lack of standardized, high-quality, and publicly available UWF-DR datasets has severely restricted the generalization ability and reliability of AI models in clinical practice. To address this, this study constructed a dataset comprising 1,630 UWF fundus images from 809 patients, which were annotated and classified by three senior ophthalmologists, for development and validation of AI system in UWF-based DR diagnosis. This dataset aims to empower researchers to train more efficient and accurate AI-assisted DR diagnosis systems based on UWF images, advancing its widespread real-world clinical applications.

DOI: https://doi.org/10.1038/s41597-026-07093-7
Sci Rep. 2026 Apr 03.

SoleFusion-Net: an explainable multimodal deep learning framework for diabetic foot syndrome classification in type II diabetes mellitus.

Mehewish Musheer Sheikh, Mamatha Balachandra, Narendra V G, Arun G Maiya.



Keywords:  Diabetic foot syndrome; Explainable Artificial Intelligence; multimodal deep learning; neuropathy classification; plantar pressure imaging

DOI:  https://doi.org/10.1038/s41598-026-42207-6
Sci Prog. 2026 Apr-Jun;109(2):109(2): 368504261440964

Interpretable machine learning in type-2 diabetes prediction in patients with depressive symptoms: Insights from mendelian randomization and physical activity.

Zhao Li, Xiaoming Qiu, Wenzhong Wu, Hyunsik Kang.

  ObjectiveThis study aimed to examine the causal relationship between depression and type-2 diabetes (T2D) and to develop a machine-learning (ML)-driven risk prediction model for T2D.MethodsThis cross-sectional study performed forward and reverse Mendelian randomization (MR) and multivariable MR (MVMR) analyses using GWAS summary data. Data from the National Health and Nutrition Examination Survey 2007-2018 were used to develop a ML-based T2D risk prediction model among individuals with depressive symptoms (n = 790). Nine ML algorithms were compared, incorporating Boruta feature selection, 10-fold cross-validation, and SHapley Additive ExPlanations (SHAP) analysis to identify and interpret key predictors. Moderation analysis was conducted to evaluate whether the triglyceride-glucose (TyG) index and physical activity (PA) influence the relationship between the predictor and the outcome.ResultsInverse variance weighted MR analysis showed that genetic liability to depression increased the risk for T2D (OR = 1.38, 95% CI: 1.01-1.90), while no significant causal effect was found in the reverse direction. MVMR analysis showed that PA was inversely related to T2D risk (OR = 0.98, 95% CI: 0.96-0.99, p = 0.001). Boruta feature selection identified age and the TyG index as significant predictors. Among nine ML algorithms, LightGBM achieved the most consistent performance, with AUROC values of 0.84 (training) and 0.75 (testing), and PRAUC values of 0.977 and 0.956, respectively. SHAP analysis confirmed that the TyG index had the highest feature importance (mean SHAP value = 0.099), followed by age (0.025). Moderation analysis identified a significant interaction between the TyG index and PA (ΔR2 = -0.007, F = 4.89, p = 0.027).ConclusionsOur MR results support a causal effect of depression on T2D risk, with the TyG index and physical activity (PA) serving as modulators. Our ML-driven prediction model provides an interpretable screening tool for T2D risk in patients with depressive symptoms.

Keywords:  depression; diabetes; machine learning; prediction; risk factors

DOI:  https://doi.org/10.1177/00368504261440964
Exp Eye Res. 2026 Mar 27. pii: S0014-4835(26)00144-2. [Epub ahead of print]267 110988

OMICS data in the diagnosis of diabetic retinopathy: A comparison between transcriptome data and DNA methylation data.

Pratheeba Jeyananthan, H V B L Chandrasiri, A G F Ashfa.

   PURPOSE: Diabetic retinopathy is a serious complication of diabetes that can damage the retina of the eye. It potentially leads to vision impairment or blindness if left untreated. Even though it can be diagnosed using dilated eye examinations, advancements in current technologies have enabled us towards more accurate diagnosis using diverse digital data sources such as medical images and omics data. However, medical images were widely used in the diagnosis of this disease using machine learning algorithms. Analysing omics data offers several advantages over image-based approaches, such as lower complexity, reduced time and less computational requirements. Moreover, omics data analysis facilitates the identification of biomarkers, which can reveal some valuable insights regarding the underlying biological mechanism of the disease.
DESIGN: Hence, this study used three different omics data, such as DNA methylation, total RNA and small RNA in the diagnosis of diabetic retinopathy using different machine learning algorithms.
METHODS: Four different feature selection algorithms were individually used with each data to select the biomarkers of the study, and the best set of features was used with different machine learning algorithms to identify the model with the highest accuracy.
RESULTS: Comparing the accuracies between models showed that 14 total RNA features selected using feature importance, along with naïve Bayes algorithms, outperformed other models with an accuracy value of 0.9625 ± 0.05.
CONCLUSIONS: Naïve Bayes algorithm using total RNA data can achieve significant performance in retinopathy diagnosis.

Keywords:  DNA methylation; Diabetic retinopathy; Diagnosis; Machine learning models; Total RNA

DOI:  https://doi.org/10.1016/j.exer.2026.110988
Neural Comput Appl. 2026 Feb;pii: 28. [Epub ahead of print]38(3):

Sequential pattern transformer (SPT): a generative and interpretable framework for predicting disease trajectories.

Mohammad Assadi Shalmani, Masoud Khani, Amirsajjad Taleban, Zihao Yi, Jennifer T Fink, Christopher E Weber, Qiang Lu, Jake Luo.

  The effective integration of artificial intelligence into clinical workflows requires models that go beyond simple prediction to generate comprehensive, explainable, and actionable disease trajectories. Addressing the limitations of opaque deep learning architectures and the noise inherent in electronic health records, we introduce the sequential pattern transformer (SPT), a novel framework that synergizes sequential pattern mining with generative transformer modeling. Using four years of inpatient data from 258,460 type 2 diabetes patients, we applied the PrefixSpan algorithm to distill noisy diagnostic histories into a curated vocabulary of 95,630 statistically validated disease progression patterns. A decoder-only transformer was trained exclusively on these evidence-based sequences to learn the temporal dynamics of disease evolution. This pattern-guided approach shifts the modeling paradigm from classification to probabilistic trajectory generation. The model achieved a robust 85.78% Top-5 accuracy, significantly outperforming a standard LSTM baseline (71.47%). Beyond predictive accuracy, the framework constructs a dynamic Disease Atlas, a branching tree structure that visualizes likely future pathways, augmented by multi-level explainable AI (XAI) including learned clinical clusters, SHAP-based feature attribution, and counterfactual simulations. Crucially, this methodology is domain-agnostic and capable of efficient fine-tuning, making it a transferable solution for adapting to diverse clinical conditions and local hospital settings. SPT thus offers a transparent, robust, and scalable framework for mapping the complex temporal dynamics of disease, bridging the gap between high-performance AI and interpretable clinical application.

Keywords:  Clinical decision support; Diabetes mellitus; Disease trajectory prediction; Explainable AI (XAI); Generative AI; Healthcare informatics; Sequential pattern mining; Transformer architecture

DOI:  https://doi.org/10.1007/s00521-025-11695-4
Diabetes Obes Metab. 2026 Mar 30.

Artificial Intelligence in Diabetes Care.

Michal Dubský, Michaela Liegertová, Robert Bém, Dominika Sojáková, Vladimíra Fejfarová, Edward B Jude.

   BACKGROUND: Diabetes care requires frequent and high-stakes decisions that must be made in the setting of substantial day-to-day physiologic variability. The growing availability of continuous glucose monitoring, connected insulin delivery devices and longitudinal electronic health record data has created an opportunity for algorithm-enabled tools that can synthesise high-frequency data, reduce cognitive burden for patients and clinicians, and support safer and more consistent decision-making.
AIM: In this review, artificial intelligence (AI) is used broadly to describe computational systems that generate predictions, recommendations or automation from clinical data.
METHODS: We distinguish between algorithmic automation and control methods that underpin many currently deployed automated insulin delivery (AID) systems and machine learning-based models, including deep learning and large language models (LLMs), that are increasingly used for pattern recognition, risk prediction and natural language applications.
RESULTS: This distinction is clinically relevant because evidence standards, safety risks and governance needs vary substantially across these categories.
CONCLUSION: This narrative review summarises current and emerging applications of AI in diabetes care with an emphasis on clinical readiness, strength of evidence and implementation considerations. We highlight established applications in AID, emerging approaches that seek greater autonomy and interoperability and newer tools such as LLMs, wearables and digital twin frameworks, focusing on where evidence is strongest, where risks are highest and what safeguards are required for responsible clinical use.

Keywords:  glycaemic control; insulin pump therapy; risk prediction; type 1 diabetes

DOI:  https://doi.org/10.1111/dom.70720
Digit Health. 2026 Jan-Dec;12:12 20552076261436274

Diabetic peripheral neuropathy identification using enface optical coherence tomography and multi-head attention deep learning algorithm.

Ying Zou, Ning Huo, Li Chen, Qincheng Qiao, Xinguo Hou.

   Objective: Diabetic peripheral neuropathy (DPN) is a common chronic complication of diabetes, but current diagnostic methods are limited by invasiveness, poor sensitivity, or subjectivity. This study aims to develop a non-invasive, reliable diagnostic tool using multimodal optical coherence tomography (OCT) images and a deep learning (DL) algorithm with multi-head attention for early DPN detection.
Methods: A multi-head attention-based DL model was constructed, with ResNet-18 as the feature extractor to fuse and classify enface OCT images from different retinal layers. A total of 3264 OCT images from 544 eyes of 435 diabetic patients were enrolled. The model was evaluated via fivefold cross-validation on the training dataset (Dataset A, n = 267) and further validated on a temporal validation dataset (Dataset B, n = 168). Single-layer contrast experiments were conducted to identify the most predictive retinal layer, and Gradient-weighted Class Activation Mapping (Grad-CAM) was used for model visualization.
Results: The proposed model achieved an average area under the curve (AUC) of 0.719 in fivefold cross-validation and an AUC of 0.721 in the temporal validation dataset. Among all retinal layers, the avascular layer provided the highest predictive value for DPN (average AUC = 0.707), with significant differences in performance compared to other layers (p < 0.05). Grad-CAM visualization revealed that photoreceptor defects were the key regions contributing to the model's classification decisions, suggesting an association between photoreceptor abnormalities and DPN. Additionally, the model outperformed individual retinal indicators (retinal nerve fiber layer thickness, superficial capillary plexus/deep capillary plexus density) whose AUCs ranged only from 0.524 to 0.565.
Conclusion: The multi-head attention-based DL model effectively identifies DPN using non-invasive OCT images, with the avascular layer providing critical information. This approach provides a promising clinically feasible early screening strategy, and photoreceptor defects may serve as a potential DPN biomarker, requiring further validation.

Keywords:  Diabetic peripheral neuropathy; algorithm; deep learning; optical coherence tomography; photoreceptor defects

DOI:  https://doi.org/10.1177/20552076261436274
Front Big Data. 2026 ;9 1676922

FunduScope: a human-centered, machine learning-based interactive tool for training junior ophthalmologists in diabetic retinopathy detection.

Sara-Jane Bittner, Michael Barz, Daniel Sonntag.

  Interpreting fundus images is an essential skill for detecting eye diseases, such as diabetic retinopathy (DR), one of the leading causes of visual impairment. However, the training of junior doctors relies on experienced ophthalmologists, who often lack the time for teaching, or on printed training materials that lack variability in examples. In this work, we present FunduScope, an interactive human-centered learning tool for training junior ophthalmologists, which is based on a pre-trained ML model for classifying DR. In a qualitative pre-study, we investigated the needs of junior doctors and identified gaps in recent learning procedures. In the main mixed-methods study, we examined the experience of 10 junior doctors with the tool and its impact on cognitive load, usability, and additional factors relevant to e-learning tools. Despite technical constraints our results confirm the potential of using an ML-based learning tool in medical education, addressing the time constraints of ophthalmologists, and providing learning independence for junior doctors. However, future work could extend the learning tool by using explainable artificial intelligence (XAI) to further support the clinical decision making of learners and exceeding the scope of this proof of concept to other ophthalmic diseases.

Keywords:  cognitive load; design thinking framework; e-learning; human-centered design; learning tool; machine learning; usability

DOI:  https://doi.org/10.3389/fdata.2026.1676922
iScience. 2026 Apr 17. 29(4): 115294

Glucose forecasting and hypoglycemia forewarning in type 1 and type 2 diabetes using deep learning.

Siyi Fang, Haowei Zhang, Die Hu, Xuefeng Yu, Zhelong Liu, Delin Ma, Weijie Xu, Fan Lin, Qiang Xie, Fang Liu, Xianlong Hu, Tangdong Ao, Dengshi Zhou, Qiang Li, Peng Zhang.

  Hypoglycemia is a major barrier to safe diabetes management. Although deep learning has been widely applied to blood glucose (BG) prediction, most studies provide limited hypoglycemia forewarning and are trained on small type 1 diabetes cohorts with restricted generalizability. We developed MT-HypoNet, a multitask neural network for real-time BG prediction and hypoglycemia forewarning from continuous glucose monitoring data. To improve detection near the hypoglycemia boundary, we introduce a statistically guided soft-label strategy. MT-HypoNet was validated on a multicenter cohort of 1,662 patients with type 1 and type 2 diabetes and prospectively evaluated in 36 perioperative patients with type 2 diabetes. In internal validation, MT-HypoNet achieved an AUC of 0.946 (95% CI: 0.946-0.947) for hypoglycemia forewarning and an RMSE of 19.84 ± 4.92 mg/dL for BG prediction. It generalized well to external datasets and maintained high prospective performance (AUC 0.966; RMSE 16.62 ± 4.01 mg/dL), supporting proactive management and improved safety.

Keywords:  health sciences; internal medicine; medical specialty; medicine

DOI:  https://doi.org/10.1016/j.isci.2026.115294
Diabetes Res Clin Pract. 2026 Mar 28. pii: S0168-8227(26)00147-6. [Epub ahead of print]235 113228

Early prediction of insulin requirement in gestational diabetes using a parsimonious LASSO model and a point-based risk score.

Domenico Greco, Salvatore Scibetta, Laura Giambanco, Vito Iannone, Luigi Calvo, Salvatore Corrao.

   AIMS: This study developed and validated a practical prediction model to forecast insulin requirements in pregnant women with gestational diabetes mellitus (GDM).
METHODS: A retrospective single-center cohort of 490 consecutive women with GDM was analyzed. The primary outcome was insulin therapy initiation based on routine clinical decisions. Candidate predictors were collected at the first visit. LASSO logistic regression with 10-fold cross-validation was used for model development, excluding underweight women (BMI < 18.5 kg/m2) and incomplete cases. Coefficients were converted into a point score (scaling factor 0.25). Discrimination was assessed using the area under the receiver operating characteristic curve (AUC), and likelihood ratios defined clinically actionable categories.
RESULTS: Insulin therapy was initiated in 11.0% of women. The derivation sample included 403 women (43 receiving insulin). Five predictors were selected: BMI category, ethnic risk background, chronic hypertension, medically assisted reproduction, and number of abnormal oral glucose tolerance test values. The model showed good discrimination (AUC 0.7755, 95% CI 0.695-0.853), excellent calibration (slope 1.02), and low Brier score (0.082). Scores ≥ 14 identified high-risk patients (LR + 11.7).
CONCLUSIONS: This LASSO-derived score using readily available variables accurately predicts insulin need in GDM, potentially facilitating risk-stratified management in community settings.

Keywords:  Clinical decision support; Gestational diabetes mellitus; Insulin therapy; Prediction model LASSO regression; Risk stratification

DOI:  https://doi.org/10.1016/j.diabres.2026.113228
Diabetes Metab Syndr Obes. 2026 ;19 590000

Machine Learning-Derived Predictive Risk Score for Prediabetes and Type 2 Diabetes Development in Parous Women.

Amélie Taschereau, Jenna Wong, Soren Harnois-Leblanc, Marie A Brunet, Myriam Doyon, Mélina Arguin, Sheryl L Rifas-Shiman, Emily Oken, Patrice Perron, Pierre-Étienne Jacques, Luigi Bouchard, Marie-France Hivert.

   Background/Aims: Pregnancy is a window of opportunity for closer links with clinical care, and to identify women at risk of chronic disease. Because of the elevated risk of type 2 diabetes (T2D) associated with gestational diabetes mellitus (GDM), most existing prediction models for post-delivery T2D focus on women with GDM, leaving many parous women without clear risk stratification. This study aimed to develop a prediction model for prediabetes or T2D risk in the general population of parous women, based on clinical pregnancy variables.
Methods: We assessed prediabetes/T2D five years after delivery in the Genetics of Glucose Regulation in Gestation and Growth (Gen3G) cohort (N=403). Using a machine-learning approach, we developed a risk prediction model from which we derived a simple, clinically usable risk index: the Gestational 4-variable Prediabetes/type 2 diabetes (G4PD) index. The G4PD index was then validated in the Project Viva cohort at three years (n=562) and seventeen years (n=541) after delivery.
Results: The G4PD index included gestational weight gain, pre-gestational body mass index, first-trimester maternal age, and a GDM variable reflecting hyperglycemia severity during pregnancy. In Gen3G, the model achieved a cross-validated estimate of the area under the receiver operating characteristic curve (ROC-AUC) of 0.696. The G4PD index achieved ROC-AUC of 0.682 in the 17-year Project Viva dataset, with similar results in the 3-year dataset. Beyond overall discrimination, the model effectively stratified women into clinically meaningful risk categories, with those in the lowest group (<2) exhibiting an expected risk of ~2% and ~15% at three and seventeen years after delivery, respectively, whereas those with the highest scores (≥7 or ≥5) expected substantially higher risks (~7% and ~37% at respective time points).
Conclusion: The G4PD index, derived from clinical pregnancy variables, moderately predicts the risk of prediabetes/T2D over several years.

Keywords:  machine-learning algorithm; prediabetes; prediction model; pregnancy; risk stratification; type 2 diabetes

DOI:  https://doi.org/10.2147/DMSO.S590000
Diabetes Metab Syndr. 2026 Mar 30. pii: S1871-4021(26)00032-9. [Epub ahead of print]20(4): 103405

Role of retinal biomarkers in diabetes detection and risk prediction: A systematic scoping review.

Soujanya Kaup, Sowrabha Bhat, Erin E Reardon, Vijayaraghavan Prathiba, Ramachandran Rajalakshmi, Rodrigo M Carrillo-Larco, Ram Jagannathan, Ranjit Mohan Anjana, Mohammed K Ali, Viswanathan Mohan, Sudeshna Sil Kar, Anant Madabhushi, Mary Beth Weber, K M Venkat Narayan.

   AIMS: This systematic scoping review was conducted to map and synthesize literature on retinal biomarkers associated with type 2 diabetes (T2D), including conventional and Artificial Intelligence(AI)-derived features, while considering ethnic/geographic diversity.
METHODS: Following PRISMA-ScR statement, seven databases were searched up to January 28, 2026 for cross-sectional and longitudinal studies in adults (≥18 years) with or without T2D. Included studies required a normoglycemic comparator and quantitative retinal assessment, with or without AI. Two reviewers independently performed title/abstract screening, full-text review and quality appraisal (using Newcastle Ottawa scale(NOS)/modified NOS); disagreements were resolved by a third reviewer. Data extraction focused on biomarker type, methodology, and population characteristics.
RESULT: Thirty-four studies (27 cross-sectional; 7 longitudinal) were included, mostly evaluating Caucasian/White [n = 18], Asian [n = 15] and Mixed/Latin American [n = 1] populations. Consistent vascular diabetes biomarkers included wider arteriolar and venular calibers. Arteriolar narrowing predicted incident diabetes in longitudinal analyses. Increased vessel tortuosity and altered fractal dimension were frequently observed. Ethnic variation was more pronounced in longitudinal studies; AI-based research was largely confined to China.
CONCLUSIONS: Retinal biomarkers show promise for early diabetes detection, but research remains geographically skewed and requires multiethnic validation to ensure clinical impact. These findings provide the biological ground truth necessary for handcrafted feature extraction in the development of interpretative artificial intelligence.

Keywords:  Artificial intelligence; Diabetes; Oculomics; Retinal biomarkers; Systematic scoping review

DOI:  https://doi.org/10.1016/j.dsx.2026.103405
Diabet Med. 2026 Mar 28. e70299

A systematic review of the use of artificial intelligence in mental health-based diabetes care: Current applications and future directions.

Vinayak Dev, Nathan S Consedine, Anna Boggiss, Anna Serlachius.

   AIMS: To map and systematise existing research on the use of artificial intelligence (AI) in mental health-based diabetes care contexts, identify trends and potential gaps in the literature, examine methodological limitations and highlight future research directions.
METHODS: The review adhered to PRISMA guidelines and was pre-registered on PROSPERO (CRD420251167053). Comprehensive searches were conducted across nine databases, including MEDLINE, EMBASE, PsycINFO and IEEE Xplore, using a Boolean strategy that combined terms related to diabetes, AI and clinical mental health. Inclusion criteria encompassed peer-reviewed, empirical, quantitative studies involving humans, diabetes contexts, mental health factors and AI-based methodologies. Screening and data extraction were performed independently by two reviewers. Forty-one studies ultimately met the inclusion criteria.
RESULTS: Research on AI in mental health-based diabetes care contexts has grown substantially since 2020. Most studies employed observational (83%) and cross-sectional (56%) designs, focused on assessment rather than intervention (88%) and targeted depression (56%). Supervised learning algorithms were most frequently used (83%); however, deep learning models achieved the highest performance. Despite technological advances, no temporal improvement in algorithmic performance was observed. Methodological limitations included limited diversity in samples and outcomes, minimal use of prospective experimental and randomised controlled trial-based designs and overreliance on supervised learning algorithms.
CONCLUSIONS: AI shows promise in addressing mental health needs in diabetes care. However, current research is narrow in scope and lacks methodological rigour in some respects. Future studies might profitably prioritise diverse populations, prospective designs, interpretability and clinical utility to enable safe, effective and equitable integration of AI into person-centred diabetes care.

Keywords:  AI; diabetes; health psychology; machine learning; mental health; psychosocial; support

DOI:  https://doi.org/10.1111/dme.70299
J Nutr. 2026 Mar 30. pii: S0022-3166(26)00161-6. [Epub ahead of print] 101512

Evaluating ChatGPT for Carbohydrate Counting Accuracy in Diabetes Management: A Precision Health Approach.

Serkan Aslan, Saniye Sözlü.

   BACKGROUND: Artificial intelligence (AI) tools have recently shown significant progress in advancing precision health. However, despite this progress, the use of AI in carbohydrate counting-a skill essential for diabetes management that presents challenges in learning and implementation-remains limited.
OBJECTIVES: This study aimed to evaluate the effectiveness of ChatGPT in creating a snack and diet plan with a certain amount of carbohydrate using carbohydrate counting.
METHODS: Study data were obtained using ChatGPT (GPT-5 version), which is freely available to the public. ChatGPT was asked to create a daily diet plan with a 45% carbohydrate content and four different snack options (standard, dairy, fruit, and whole grains) with 15 g and 30 g of carbohydrates. Carbohydrate and energy values of the foods were calculated using a nutrition database.
RESULTS: For snack prompts, carbohydrate values did not significantly differ from the predefined 15 g and 30 g targets in most categories (p > 0.05). Significant differences were found in absolute deviation values in the 15 g prompt-standard group. In the 2000-kcal daily diet plan, the mean carbohydrate percentage (45%) and gram value (225 g) were achieved; however, total energy values significantly exceeded the 2000-kcal target (p = 0.008).
CONCLUSION: Although ChatGPT demonstrated high accuracy in approximating predefined carbohydrate targets at the mean level, deviations in total energy and variability across prompts suggest challenges in simultaneously satisfying multiple nutritional constraints. These findings highlight both the potential and the limitations of ChatGPT in carbohydrate counting and support the need for expert oversight in clinical applications.

Keywords:  ChatGPT; artificial intelligence; carbohydrate counting; diabetes; precision nutrition

DOI:  https://doi.org/10.1016/j.tjnut.2026.101512
medRxiv. 2026 Feb 24. pii: 2026.02.23.26346903. [Epub ahead of print]

Multimodal AI fuses proteomic and EHR data for rational prioritization of protein biomarkers in diabetic retinopathy.

Jonathan B Lin, Samson J Mataraso, Madhumeeta Chadha, Gabriel Velez, Prithvi Mruthyunjaya, Nima Aghaeepour, Vinit B Mahajan.

Purpose: There is a need for novel therapies for diabetic retinopathy (DR) because existing therapies treat only certain features of DR and do not work optimally for all patients. While proteomic studies provide insight into disease pathobiology, they are often limited to small sample sizes due to high costs, limiting their generalizability and reproducibility. Moreover, they often yield lists of tens to hundreds of proteins with differential expression, making it difficult to prioritize the most biologically relevant biomarkers beyond using arbitrary fold-change and false-detection rate cutoffs. Here, we applied a two-stage multimodal AI approach: first, we integrated EHR and proteomics data to rationally prioritize candidate protein biomarkers and, next, validated these biomarkers in an independent cohort. These protein biomarkers of DR are rooted in the EHR data and thereby more likely to be biological drivers of disease.
Methods: We obtained EHR data from a large number of patients with and without DR (N=319,997) from the STARR-OMOP database and obtained aqueous humor liquid biopsies from a subset of these patients (N=101) for high-resolution proteomic profiling. We developed C linical and O mics M ulti-Modal Analysis E nhanced with T ransfer Learning (COMET) to perform integrated analysis of proteomics and all available EHR data to identify protein biomarkers of DR. The model was trained in two phases: first, it was pretrained using patients with EHR data alone (N=319,896), and then, it was fine tuned using patients with both EHR and proteomics data (N=101), allowing it to learn both clinical and molecular features associated with DR. Findings from COMET were then validated with liquid biopsies from an independent, validation cohort (N=164).
Results: t-distributed stochastic neighbor embedding (t-SNE) analysis of EHR and proteomics data identified proteins clustering with related EHR features. Levels of STX3 and NOTCH2, proteins involved in retinal function, were correlated with a diagnosis of macular edema, a record of a visual field exam, and a prescription for latanoprost, highlighting protein-EHR alignment. The pretrained, multimodal COMET model was superior (AUROC=0.98, AUPRC=0.91) compared to models generated using either EHR or proteomics data alone or without pretraining (AUROC: 0.76 to 0.92; AUPRC: 0.47 to 0.74). The proteins SERPINE1, QPCT, AKR1C2, IL2RB, and SRSF6 were prioritized by the COMET model compared to the models without pretraining, supporting their potential role in DR pathobiology, and were subsequently validated in an independent cohort.
Conclusion: We used multimodal AI to prioritize protein biomarkers of DR that are most strongly linked to EHR elements, as well as identifying other protein biomarkers associated with disease features like diabetic macular edema. These findings serve as a foundation for future mechanistic studies and highlight the synergistic value of using multimodal AI to fuse EHR and proteomics data for enhanced proteomics analysis.

DOI: https://doi.org/10.64898/2026.02.23.26346903
Front Cell Dev Biol. 2026 ;14 1763178

Identification of mitochondria-related key genes in type 2 diabetes mellitus and elucidation of the Zhimu-Huangbai herb Pair's mechanism: an integrated approach of bioinformatics, machine learning, and experimental validation.

Sufang Zhang, Minghe Gu, Baoting Chen, Yiming Liu, Aihua Lin.

   Background: Mitochondrial injury plays a critical role in type 2 diabetes mellitus (T2DM) pathogenesis by impairing cellular energy metabolism and insulin sensitivity. The Zhimu-Huangbai herb pair (ZB), a classic Traditional Chinese Medicine formulation composed of Anemarrhena asphodeloides and Phellodendron chinense, has shown efficacy in T2DM, but its molecular mechanisms remain unclear. In this study, we aimed to identify crucial mitochondrial related genes of type 2 diabetes and the potential mechanism of ZB.
Methods: Gene expression datasets for T2DM (GSE76894, GSE25724, and GSE38642) were retrieved from the GEO database. Intersection targets of ZB herb pair and T2DM were identified by screening multiple databases, including the TCMSP and HERB. Mitochondrial function-related genes were obtained from human mitochondria-associated databases. WGCNA was employed to identify differentially expressed genes, which were then intersected with bioactive compound-target genes and mitochondrial-related genes to construct a PPI network. GO and KEGG enrichment analyses were subsequently performed. Four machine learning algorithms-SVM-RFE, RF, GLM, and XGB-were applied to screen feature genes and establish diagnostic models. Furthermore, the correlations between feature targets and immune cell infiltration were analyzed, single-gene GSEA was conducted, and molecular docking was performed to investigate the interactions between feature targets and bioactive constituents of ZB. For experimental validation, INS-1 cells were divided into six groups: the control group, model group, metformin group, and low-, medium-, and high-dose ZB groups. Cell viability, apoptosis, ROS levels, mitochondrial membrane potential, and mitochondrial morphology and function were assessed. Western blot analysis was performed to evaluate the expression of mitochondria-related genes (BCAT2, CASP8, EPHX2, and UCP2) and components of the AMPK-SIRT1-PGC-1α signaling pathway.
Results: A total of eight mitochondria-related differentially expressed genes associated with ZB treatment of T2DM were identified. GO analysis revealed enrichment in multiple biological processes, including response to nutrient levels; cellular components, such as pore complex; and molecular functions, including toxic substance binding. KEGG pathway analysis indicated significant enrichment in pathways including apoptosis, p53 signaling pathway, and necroptosis. Three key genes-BCAT2, CASP8, and EPHX2-were screened through machine learning algorithms, and the constructed T2DM diagnostic models all exhibited area under the curve (AUC) values greater than 0.7, indicating satisfactory discriminative performance. Immune infiltration analysis revealed that all three key genes were significantly correlated with immune cell populations. Molecular docking results demonstrated that the three key genes exhibited strong binding affinities (≤-5.0 kcal/mol) for their corresponding bioactive compounds derived from ZB, with the exception of the CASP8-nicotinamide combination. Experimental validation showed that ZB significantly enhanced the viability of INS-1 cells subjected to high-glucose and high-lipid conditions, inhibited apoptosis, reduced intracellular ROS generation, and ameliorated mitochondrial membrane potential, mitochondrial morphology, and respiratory function. Concurrently, the protein expression levels of UCP2 and BCAT2 were markedly upregulated, whereas those of CASP8 and EPHX2 were significantly downregulated. Additionally, ZB treatment upregulated the p-AMPK/AMPK ratio as well as the expression of SIRT1 and PGC-1α.
Conclusion: The diagnostic model featuring genes BCAT2, CASP8, and EPHX2 provides new insights for T2DM diagnosis and treatment. ZB's therapeutic mechanism involves regulating mitochondrial-related genes (BCAT2, CASP8, EPHX2, UCP2) and activating the AMPK-SIRT1-PGC-1α pathway, thereby improving mitochondrial morphology and function, reducing oxidative damage, and enhancing energy metabolism.

Keywords:  bioinformatics; energy metabolism; machine learning; mitochondria; type 2 diabetes mellitus; zhimu-huangbai

DOI:  https://doi.org/10.3389/fcell.2026.1763178
J Diabetes Sci Technol. 2026 Mar 29. 19322968261434035

A Deep Learning Prognostic Model for Diabetes Patients Using Bilateral Fundus Imaging.

Yen-Ming Lai, I-Fang Chung, Shang-Hsuan Huang, Wayne Huey-Herng Sheu, Kung-Hao Liang, Harn-Shen Chen.

   AIMS: Diabetes mellitus shortens life expectancy, driven primarily by premature mortality from vascular complications. Mortality models for intensive care units are well established, whereas outpatient mortality prediction remains challenging. To empower proactive, life-extending care in aging societies, we developed a prognostic model that utilizes routine bilateral color fundus photography for identifying high-mortality-risk patients years before irreversible complications, enabling timely interventions that can substantially improve survival.
METHODS: We analyzed 19 029 persons with diabetes who received routine color fundoscopy examinations. We adapted deep learning architectures to align with Cox proportional hazards modeling of mortality events, as opposed to simple classification.
RESULTS: Our image-trained mortality model stratified the testing cohort into quartiles exhibiting escalating hazard ratios (HRs) of 1.67 (95% confidence interval [CI]: 0.80-3.47), 3.02 (1.55-5.88), and 7.19 (3.91-13.25) relative to the lowest-risk group. The five-year survival rate of the highest-risk quartile was lower than 80%. For comparison, we evaluated a regulatory-approved commercial ophthalmic system, which generated HRs of 1.48, 2.02, and 3.11 across its strata. When evaluated in strata of equivalent sizes to those used by the commercial system, our system demonstrated stronger separation than the commercial system, yielding HRs of 3.64 (95% CI: 3.18-4.16), 4.98 (4.08-6.07), and 6.89 (6.01-7.89). The top 8% patients identified by our model exhibited a five-year survival rate <70%.
CONCLUSION: We developed a survival-aware, image-based model that can predict mortality risk directly from routine fundus photography, achieving prognostic discrimination which facilitates diabetes management.

Keywords:  color fundus; diabetes mortality; end-to-end survival learning; prognostic model

DOI:  https://doi.org/10.1177/19322968261434035
Cureus. 2026 Feb;18(2): e104417

Integrating a Smart Sensor Chip and AI Predictive Analytics Into the Sehhaty App to Enhance Diabetes Management in Saudi Arabia.

Abdullah F ALqunisi.

  Diabetes is a major public health challenge in the Kingdom of Saudi Arabia. A substantial proportion of adults are affected, placing significant pressure on the healthcare system. Although digital health initiatives have expanded in recent years, patients continue to encounter barriers to adopting mobile health (mHealth) technologies, including technical limitations, usability concerns, and privacy issues. This article proposes a comprehensive digital health solution for diabetes management that enhances the national health application Sehhaty by integrating two complementary technologies: a smart sensor chip (SSC) for continuous physiological monitoring and AI-based predictive analytics (AIPA) for forecasting glycemic trends. The aim is to strengthen proactive and personalized diabetes care. An Agile development framework consisting of two sprints is proposed. The first sprint integrates a wearable SSC into the Sehhaty ecosystem to transmit real-time glucose and vital sign data through secure wireless communication. The second sprint develops AIPA using machine-learning models to analyze data patterns and predict glycemic fluctuations. System requirements were identified through stakeholder engagement, and the architecture includes a secure cloud backend, structured data flow, and user-centered interface design. The integration of SSC and AIPA into Sehhaty could enhance diabetes management by enabling continuous monitoring, personalized alerts, and earlier intervention. The system design prioritizes reliability, user-centered usability, and data privacy safeguards to address common barriers to digital health adoption. Consideration of perceived usefulness, ease of use, trust, and accessibility informed the development strategy. The proposed SSC-AIPA framework has the potential to transform Sehhaty into an advanced diabetes management platform that supports early detection, predictive insights, and individualized care. Future work will include prototyping, usability evaluation, and clinical validation.

Keywords:  diabetes management; mobile health; predictive analytics; saudi arabia; sehhaty app; smart sensor chip

DOI:  https://doi.org/10.7759/cureus.104417
Discov Anal. 2026 ;4(1): 10

Flexible statistical approaches for modeling nonlinear relationships in diabetes prediction using splines, Bayesian kernel regression and Bayesian regression trees.

Thimani Dananjana Ranathungage, Harsha Blumer, Saman Muthukumarana.

  Modeling nonlinear relationships is a fundamental challenge in statistical analysis, particularly when predictors exhibit complex and interacting effects on outcomes. This study compares three flexible methods for capturing such structures: restricted cubic spline regression (RCS), Bayesian kernel machine regression (BKMR), and Bayesian additive regression trees (BART). RCS enables explicit modeling of nonlinear associations via spline basis functions, BKMR leverages kernel functions within a Bayesian framework to capture nonlinear and non-additive effects, and BART provides a nonparametric ensemble approach that flexibly accommodates interactions and nonlinearities without prior specification. To demonstrate the utility of these methods, we apply them to the Pima Indians diabetes dataset, consisting of 768 observations of women at high risk of type II diabetes. After data preprocessing, including imputation and outlier handling, each method was fitted and evaluated using six performance measures. RCS and BKMR identified glucose, insulin, age, and skin thickness as significant predictors, while BART yielded the best predictive performance (AUC [Formula: see text] 97%). Predictor-response functions were used to enhance clinical interpretability. These findings illustrate that a methodology capable of capturing nonlinear effects can substantially improve prediction accuracy in epidemiological studies.

Keywords:  Bayesian additive regression trees; Bayesian kernel machine regression; Nonlinear effects; Restricted cubic splines regression

DOI:  https://doi.org/10.1007/s44257-026-00057-6