bims-aukdir 2025-11-23 papers

bims-aukdir

Biomed News

on Automated knowledge discovery in diabetes research

Issue of 2025–11–23
seventeen papers selected by
Mott Given

Evaluation of the degree of agreement in the diagnosis of diabetic retinopathy between ophthalmologists and EyeArt®.
Role of artificial intelligence-enabled hand-held fundus camera for community-based diabetic retinopathy screening.
DRSplus retinographer in patients with diabetes: a comparison study by consensus grading, Reading Center and Artificial Intelligence.
Reach and implementation of human and AI-assisted diabetic retinopathy screening models in primary healthcare settings in India.
The use of machine learning to predict pharmacological therapy in gestational diabetes: A scoping review.
Integration of artificial intelligence and wearable technology in the management of diabetes and prediabetes.
Lightweight Sequential Transformers for Blood Glucose Level Prediction in Type-1 Diabetes.
Use of continuous glucose monitoring systems in pediatric patients in the perioperative environment: Challenges and machine learning opportunities.
Real-World Prospective Validation and Economic Evaluation of Deep Learning-Based Diabetic Retinopathy Detection From Fundus Photographs: A Systematic Review and Meta-analysis.
Data Augmentation Via Digital Twins to Develop Personalized Deep Learning Glucose Prediction Algorithms for Type 1 Diabetes in Poor Data Context.
Plasma metabolomics disentangles T2DM- and CAD-specific dysmetabolism and identifies potential biomarkers for CAD risk escalation in diabetic patients.
Development and External Validation of a Machine Learning-Based Model for Predicting Heart Failure Risk in Type 2 Diabetes.
Aetiological clustering of newly diagnosed type 2 diabetes using machine learning: a retrospective cross-sectional study in Dubai, UAE.
DeepGCGR: an interpretable two-layer deep learning model for the discovery of GCGR-activating compounds.
Artificial intelligence-driven clinical decision support systems to assist healthcare professionals and people with diabetes in Europe at the point of care: a Delphi-based consensus roadmap.
My diabetes care: an AI-based mobile app with conversational agent for type 2 diabetes self-management.
Patterns of discussion on neuroticism and self-management behaviors in type 2 diabetes: a scoping review using machine learning-assisted text mining.

Int J Retina Vitreous. 2025 Nov 19. 11(1): 125

Evaluation of the degree of agreement in the diagnosis of diabetic retinopathy between ophthalmologists and EyeArt®.

Isabel Inmaculada Guedes Guedes, Pedro Saavedra Santana, Francisco Cabrera López, Ángel Ramos Macías, Ángel Ramos de Miguel, Ayoze González Hernández.

   OBJECTIVE OR PURPOSE: To evaluate the diagnostic performance and agreement of the EyeArt® Artificial Intelligence (AI) system for detecting Diabetic Retinopathy (DR), comparing its results with ophthalmologists' assessments in a regional screening program.
DESIGN: Cross-sectional observational study.
SUBJECTS, PARTICIPANTS, AND/OR CONTROLS: A total of 498 diabetic patients aged 18 years or older were enrolled between June and September 2023 through the Retisalud DR screening program in the Canary Islands. No separate control group was included.
METHODS: All participants underwent non-mydriatic fundus photography using the TRC-NW400 camera. Retinal images were analyzed by the EyeArt® AI system (version 2.1.0), and results were compared with assessments by ophthalmologists based on the International Clinical Diabetic Retinopathy scale (ICDR). Agreement was quantified using Cohen's kappa coefficient. Additionally, mixed-effects logistic regression was used to explore associations between DR and clinical risk factors.
MAIN OUTCOME MEASURES: Sensitivity, specificity, and agreement (Cohen's kappa) of the AI system compared to clinical diagnosis; predictors of DR such as age, diabetes duration, presence of Diabetic Macular Edema (DME), and central retinal thickness (CRT-OCT).
RESULTS: The EyeArt® system achieved a binocular sensitivity of 100% (95% CI: 98.1-100) and a specificity of 93.5% (95% CI: 90.2-96.0). Agreement with ophthalmologist grading was excellent, with kappa values of 0.966 (right eye) and 0.978 (left eye). Younger age, longer diabetes duration, DME presence, and higher CRT were significantly associated with DR diagnosis.
CONCLUSIONS: The EyeArt® AI system showed excellent diagnostic accuracy and strong agreement with clinical evaluations in DR screening. Nonetheless, its tendency to overestimate DR severity indicates the need for further refinement of its grading algorithm. These findings support the potential integration of AI systems into large-scale DR screening programs, pending further validation.

Keywords:  Artificial intelligence; Artificial intelligence detection of diabetic retinopathy; Automated diabetic retinopathy; Automated retinal image; Diabetic retinopathy

DOI:  https://doi.org/10.1186/s40942-025-00748-4
Med J Armed Forces India. 2025 Nov-Dec;81(6):81(6): 665-671

Role of artificial intelligence-enabled hand-held fundus camera for community-based diabetic retinopathy screening.

Vijay K Sharma, Srishti Khullar, Prabhjot Singh, Vikas Ambiya, Ashok Kumar, Anuroop N, Gaurav Kapoor, Preeti Rk.

   Background: This study aimed to assess the diagnostic accuracy of an artificial intelligence (AI) system integrated with a portable handheld fundus camera for the detection of diabetic retinopathy (DR) in a community-based screening program.
Methods: A DR screening camp was organized at a tertiary care hospital in India. A cohort of 261 patients with diabetes was screened using a nonmydriatic handheld fundus camera. Retinal images were graded by specialists and compared with the AI system's output. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the receiver operating characteristic curve (AUC-ROC) were calculated. Subgroup analyses based on image quality was performed.
Results: Of the 261 patients screened, 253 had available retinal images, and 243 had gradable images. The AI system achieved a sensitivity of 85.29%, specificity of 99.04%, PPV of 93.55%, and NPV of 97.64% for detecting referable DR. The AUC-ROC was 0.93. The AI system's performance remained robust across all image-quality categories. The AI system showed strong agreement with human graders (κ = 0.86). However, it failed to identify certain non-DR pathologies detected by human graders.
Conclusions: The AI system integrated with a portable handheld fundus camera demonstrated high diagnostic accuracy for referable DR detection in a community-based screening setting. This technology shows promise for expanding DR-screening coverage in resource-limited settings.

Keywords:  Artificial intelligence; Diabetic retinopathy; Diagnostic imaging; Mass screening; Telemedicine

DOI:  https://doi.org/10.1016/j.mjafi.2024.09.008
Retina. 2025 Nov 12.

DRSplus retinographer in patients with diabetes: a comparison study by consensus grading, Reading Center and Artificial Intelligence.

Parravano Mariacristina, Costanzo Eliana, De Geronimo Daniele, Varano Monica, Rossi Tommaso, Rui Chiara Eng, Gazzina Silvia Eng, Querques Giuseppe.

   PURPOSE: To evaluate the gradable rate of the retinal images acquired with DRSplus retinographer in patients with diabetes and to estimate the diabetic retinopathy (DR) severity, comparing different methods of analysis and grading.
METHODS: Prospective, cross-sectional, observational study. A mosaic of overlapped retinal images in non-mydriatic condition was acquired, evaluating the gradable images rate at both image and eye-levels, from the human-study team (consensus grading), Reading Center (RC) and an Artificial Intelligence (AI) system, respectively. The DR severity was graded by the RC and the AI.
RESULTS: 844 images of 422 eyes from 224 patients were included. The gradable rate was 87.4%, 97.4% and 96.9% at image-level and was 81.3%, 97.4% and 95.0% at eye-level, according to the consensus grading, the RC and the AI, respectively. AI sensitivity for detecting referable DR was 97.3%, the specificity was 80.4% and the accuracy was 86.6%.
CONCLUSIONS: DRSplus retinographer allowed to image the retina in non-mydriatic condition in more than 80% of eyes, with a higher gradable rate with RC and AI in comparison with consensus grading. The estimation of the DR severity showed a high sensitivity for referral cases for AI system, indicating the potentialities of telemedicine-based tool for the DR screening.

Keywords:  Diabetic retinopathy; gradeability analysis; non-mydriatic fundus camera

DOI:  https://doi.org/10.1097/IAE.0000000000004734
Sci Rep. 2025 Nov 21. 15(1): 41355

Reach and implementation of human and AI-assisted diabetic retinopathy screening models in primary healthcare settings in India.

Anshul Chauhan, Luke Vale, Ankita Kankaria, Vishali Gupta, Mandeep Singh, Gagandeep Kaur, Sonam Kumar, Mitesh Yadav, Neha N, Basavaraj Tigari, Sanjay Bhadada, Mona Duggal.

  Diabetic retinopathy (DR) is a leading cause of preventable vision loss. While DR screening is critical, evidence on the reach and implementation of different screening models in primary healthcare settings is limited. This study evaluated the reach and implementation of DRS models in northern India using the RE-AIM framework. A pragmatic three-arm observational study was conducted between February 2023 and January 2024 in Block Boothgarh, a rural block in District Mohali, Punjab, comprising 30 villages with an estimated 120,000 residents. Household line listing was performed to identify individuals aged 30 years or older with diabetes. Participants (n = 600) were equally allocated to three screening models: facility-based screening at Health and Wellness Centres (HWC) by non-ophthalmologists, community-based AI-assisted screening at home, and standard care. Reach and implementation were assessed through quantitative data, field observations, and qualitative interviews with healthcare providers. Refusal for screening was higher in facility-based screening (40%, 135/340) and lower in community-based screening (13%, 31/240). Older individuals were more likely to decline participation, with a mean age of 62.0 years for males and 60.3 years for females. Reported barriers included existing medical conditions, mobility limitations, perceived good eye health, travel distance, and transportation difficulties. Concerns regarding long-term medication adherence also reduced uptake. Technical issues, including power outages, hardware or software malfunctions, suboptimal image quality, and lack of cooperation, further declined implementation. Adaptations, including the use of backup power generators, on-site troubleshooting, and provision of transport support, mitigated these barriers and improved overall implementation fidelity. Assessing reach is essential for the success of public health interventions. Using the RE-AIM framework, this study identified key barriers and adaptive strategies in DRS, enhancing both reach and implementation within primary healthcare settings. These findings can inform the integration of DRS models into comparable resource-constrained contexts, thereby improving overall effectiveness.Clinical Trial Registry of India (CTRI): 2022/10/046283.

Keywords:  Artificial intelligence; Diabetic retinopathy screening; Implementation; Pragmatic trial; RE-AIM; Reach

DOI:  https://doi.org/10.1038/s41598-025-25402-9
Diabet Med. 2025 Nov 18. e70171

The use of machine learning to predict pharmacological therapy in gestational diabetes: A scoping review.

Jasmine R Kirkwood, Natasha Galloway, Robert S Lindsay, Areti Manataki, Deborah J Wake, Rebecca M Reynolds.

   AIMS: Early identification of pharmacological therapy for gestational diabetes mellitus (GDM), a common pregnancy complication, through machine learning could allow for better therapeutic strategies and improved treatment efficiency. This scoping review aimed to comprehensively review the machine learning models used to predict the need for pharmacological therapy in GDM.
METHODS: Four electronic databases-Embase, Medline, IEEE Xplore and Web of Science-were searched for publications between 1 July 2007 and 31 August 2024. Studies predicting pharmacological therapy for GDM using machine learning were included. The Joanna Briggs Institute and PRISMA-ScR checklist was followed, and the Prediction model Risk Of Bias ASsessment Tool (PROBAST) was used to assess quality.
RESULTS: Included were 17 studies presenting 44 models, 61.4% (27/44) predicted any pharmacological therapy use and 38.6% (17/44) predicted insulin use alone. All were binary classifiers, and logistic regression was typically used. The overall area under the receiver operating curve had a median of 0.75. Common clinical variables were found to be predictors, such as history of GDM, gestational week at GDM diagnosis, pregestational body mass index, maternal age, HbA1c, fasting and 1 h glucose from 75 g oral glucose tolerance test. Though 65.9% of models were validated, there was a lack of external validation. There was no evidence of clinical application of the models.
CONCLUSION: Logistic regression with common clinical variables was often used to predict pharmacological therapy for GDM. Few models were externally validated or clinically applicable.

Keywords:  gestational diabetes mellitus (GDM); insulin; machine learning; oral agents; pharmacological therapy; prediction algorithms

DOI:  https://doi.org/10.1111/dme.70171
NPJ Digit Med. 2025 Nov 18. 8(1): 687

Integration of artificial intelligence and wearable technology in the management of diabetes and prediabetes.

Raphael A Fraser, Rebekah J Walker, Jennifer A Campbell, Obinna Ekwunife, Leonard E Egede.

Artificial intelligence and wearable technology are increasingly used in healthcare and hold significant potential for improving the management of diabetes. Wearable devices enable continuous monitoring and real-time data collection, supporting AI-driven personalized interventions. This systematic review evaluated peer-reviewed studies that examined the integration of AI and wearable technology in diabetes management, with a focus on clinical and self-management outcomes. Sixty studies were included following a review of over 5000 records. AI models paired with wearable devices showed promise in glycemic monitoring, adaptive insulin management, and predicting diabetes-related events. Continuous glucose monitors and other wearables also enhanced self-management and informed clinical decision-making. However, key challenges persist, including limited demographic diversity, variable data quality, a lack of standardized benchmarks for evaluating AI performance, and limited interpretability of complex models. Future research should prioritize improving model transparency, addressing demographic disparities, and establishing clear benchmarks to support equitable and effective implementation in diabetes care.

DOI: https://doi.org/10.1038/s41746-025-02036-9
IEEE J Biomed Health Inform. 2025 Nov 17. PP

Lightweight Sequential Transformers for Blood Glucose Level Prediction in Type-1 Diabetes.

Mirko Paolo Barbato, Giorgia Rigamonti, Davide Marelli, Paolo Napoletano.

Type 1 Diabetes (T1D) affects millions worldwide, requiring continuous monitoring to prevent severe hypo- and hyperglycemic events. While continuous glucose monitoring has improved blood glucose management, deploying predictive models on wearable devices remains challenging due to computational and memory constraints. To address this, we propose a novel Lightweight Sequential Transformer model designed for blood glucose prediction in T1D. By integrating the strengths of Transformers' attention mechanisms and the sequential processing of recurrent neural networks, our architecture captures long-term dependencies while maintaining computational efficiency. The model is optimized for deployment on resource-constrained edge devices and incorporates a balanced loss function to handle the inherent data imbalance in hypo- and hyperglycemic events. Experiments on two benchmark datasets, OhioT1DM and DiaTrend, demonstrate that the proposed model outperforms state-of-the-art methods in predicting glucose levels and detecting adverse events. This work fills the gap between high-performance modeling and practical deployment, providing a reliable and efficient T1D management solution.

DOI: https://doi.org/10.1109/JBHI.2025.3633194
World J Clin Pediatr. 2025 Dec 09. 14(4): 107127

Use of continuous glucose monitoring systems in pediatric patients in the perioperative environment: Challenges and machine learning opportunities.

Tara Doherty, Ashley Kelley, Elizabeth Kim, Irim Salik.

  Pediatric type 1 diabetes (T1D) is a lifelong condition requiring meticulous glucose management to prevent acute and chronic complications. Conventional management of diabetic patients does not allow for continuous monitoring of glucose trends, and can place patients at risk for hypo- and hyperglycemia. Continuous glucose monitors (CGMs) have emerged as a mainstay for pediatric diabetic care and are continuing to advance treatment by providing real-time blood glucose (BG) data, with trend analysis aided by machine learning (ML) algorithms. These predictive analytics serve to prevent against dangerous BG variations in the perioperative environment for fasted children undergoing surgical stress. Integration of CGM data into electronic health records (EHR) is essential, as it establishes a foundation for future technologic interfaces with artificial intelligence (AI). Challenges in perioperative CGM implementation include equitable device access, protection of patient privacy and data accuracy, ensuring institution of standardized protocols, and financing the cumbersome healthcare costs associated with staff training and technology platforms. This paper advocates for implementation of CGM data into the EHR utilizing multiple facets of AI/ML algorithms.

Keywords:  Artificial intelligence; Continuous glucose monitor; Continuous glucose monitoring system; Electronic health records; Type 1 diabetes mellitus

DOI:  https://doi.org/10.5409/wjcp.v14.i4.107127
Diabetes Care. 2025 Nov 19. pii: dc251493. [Epub ahead of print]

Real-World Prospective Validation and Economic Evaluation of Deep Learning-Based Diabetic Retinopathy Detection From Fundus Photographs: A Systematic Review and Meta-analysis.

An Ran Ran, Jennifer Li Ding, Ziqi Tang, Ching Lam, Truong X Nguyen, Jiaying Zhou, Shuyi Zhang, Danqi Fang, Dawei Yang, Vincent Ng, Duoru Lin, Haotian Lin, Clement C Tham, Carmen K M Chan, Simon K H Szeto, Tien Y Wong, Sobha Sivaprasad, Carol Y Cheung.

BACKGROUND: Deep learning (DL) has shown promise in delivering diagnostic and economic benefits for detecting diabetic retinopathy (DR) from fundus photographs (FPs). However, evidence synthesis of model validation in prospective, real-world settings remains limited.
PURPOSE: To assess the feasibility of implementing DL-DR systems using FPs across different countries by synthesizing prospective validation and economic evidence.
DATA SOURCES: Five databases were searched until 13 August 2025.
STUDY SELECTION: Studies prospectively assessing diagnostic performance and/or studies conducting economic analyses of DL-DR systems using FPs were selected.
DATA EXTRACTION: Characteristics of all studies, performance parameters of prospective validation studies, and economic outcomes of economic analysis studies were extracted.
DATA SYNTHESIS: Forty-seven studies were included in the meta-analysis. The pooled performance was the highest in detecting vision-threatening DR (area under the receiver operating characteristic curve [AUROC] 0.974), followed by any DR (AUROC 0.965), then referable DR (RDR) (AUROC 0.959). Study region, clinical pathway, mydriasis, image quality control, sample size, grading criteria, reference standard, and model architecture significantly affected model performance in RDR detection. Fifteen studies were included in the economic commentary, showing that DL-based DR screening was cost-effective in high-income countries, whereas results in middle-income countries were mixed, depending on compliance rates, glycemic control, and initial costs.
LIMITATIONS: A paucity of studies assessing multiple severities of DR or diabetic macular edema restricted our ability to perform subgroup analyses. Insights into low-income countries were limited by a lack of studies in these regions.
CONCLUSIONS: DL-DR systems using FPs had high discriminative performance in prospective real-world settings and hold promise to improve cost-effectiveness, especially in high-income countries.

DOI: https://doi.org/10.2337/dc25-1493
IEEE Trans Biomed Eng. 2025 Nov 21. PP

Data Augmentation Via Digital Twins to Develop Personalized Deep Learning Glucose Prediction Algorithms for Type 1 Diabetes in Poor Data Context.

Francesco Prendin, Andrea Facchinetti, Giacomo Cappon.

OBJECTIVE: Accurately predicting glucose levels is essential for effectively managing type 1 diabetes (T1D), a chronic condition in which the body cannot produce insulin. Although deep learning approaches have shown promise, their training requires extensive datasets that capture a wide range of physiological and behavioral variations. However, obtaining such datasets can be challenging and impractical, especially when their collection demands significant patient effort. To overcome this limitation, we propose a data augmentation strategy that leverages digital twins of individuals with T1D (DT-T1D) to generate personalized synthetic data mirroring real-world glucose-insulin dynamics.
METHODS: ReplayBG, an open-source tool for creating DT-T1D, was adapted to develop a two-steps strategy: first, generating DT-T1D from retrospective patient data; then, using DT-T1D with new inputs, to simulate synthetic, patient-specific data. The practical impact of this approach is demonstrated in a case study where personalized deep networks were developed to predict glucose levels. Models were trained on an open-source dataset from 12 patients, using either the original data or a combination of the original and synthetic data.
RESULTS: Integrating synthetic data into the training process consistently enhances model performance. Moreover, models trained on synthetic data combined with only a small fraction of the original dataset achieve results comparable to those obtained from the full, unaugmented dataset.
CONCLUSION: Leveraging DT-T1D to generate personalized synthetic data mitigates data scarcity and enhances deep learning model performance for accurate glucose prediction.
SIGNIFICANCE: This work highlights the potential of digital twin-driven data augmentation to tackle data scarcity and develop robust, personalized predictive models for T1D management.

DOI: https://doi.org/10.1109/TBME.2025.3635264
Cardiovasc Diabetol. 2025 Nov 19. 24(1): 440

Plasma metabolomics disentangles T2DM- and CAD-specific dysmetabolism and identifies potential biomarkers for CAD risk escalation in diabetic patients.

Ting Hu, Wen Zhang, Xian Ding, Feifei Han, Zhuoling An.

   BACKGROUND: Type 2 diabetes mellitus (T2DM) is a major driver of coronary artery disease (CAD). Prior studies often conflate T2DM- and CAD-specific metabolic alterations, limiting insights into CAD pathogenesis in T2DM. This study aimed to distinguish CAD-unique signatures from T2DM-specific dysmetabolism, and to identify potential metabolic biomarkers for CAD risk escalation in T2DM patients.
METHODS: We performed an untargeted plasma metabolomic study with 123 healthy controls (HCs), 50 T2DM patients without CAD, and 155 T2DM patients with CAD. T2DM_CAD was defined as T2DM diagnosed at least 5 years prior to CAD, with coronary angiography-confirmed stenosis (> 30%) in major coronary arteries. Differential metabolites were identified via intergroup comparisons, with T2DM-specific and CAD-specific signatures distinguished based on unique expression patterns. Machine learning models were developed to evaluate the discriminatory performance of these metabolites for CAD.
RESULTS: Plasma metabolomic profiling identified distinct metabolic patterns across the three cohorts. Metabolites specific to T2DM were enriched in carbohydrates and certain lipid species, reflecting disturbances in glucose and lipid metabolism. CAD-specific metabolites were predominantly lipids and organic acids, with notable involvement in amino acid and fatty acid metabolic pathways. Several metabolites changed progressively from HCs through T2DM to T2DM_CAD, reflecting advancing metabolic dysregulation, whereas others showed opposing trends, suggesting compensatory or protective adaptations. Integration of key metabolites with clinical parameters in machine learning models effectively distinguished between study groups, demonstrating promising performance for CAD risk assessment in T2DM patients.
CONCLUSIONS: These findings disentangle T2DM- and CAD-specific metabolic disturbances and identify escalation/de-escalation features of CAD risk in diabetic patients, which are potential candidates for future risk stratification pending validation.

Keywords:  Biomarkers; Coronary artery disease; Machine learning; Metabolomics; Type 2 diabetes mellitus

DOI:  https://doi.org/10.1186/s12933-025-02981-5
Diabetes Metab Syndr Obes. 2025 ;18 4177-4191

Development and External Validation of a Machine Learning-Based Model for Predicting Heart Failure Risk in Type 2 Diabetes.

Yuqing Liu, Ping Wang, Min Wang, Yan Chen, Sania Martin Kasyanju, Yuhong Yang, Tao Yang, Li Peng, Min Sun.

   Background: Heart failure (HF) is a severe and common complication of type 2 diabetes mellitus (T2DM), associated with increased morbidity and mortality. Although the biomarker NT-proBNP, at a cut-off value of 125 pg/mL, has demonstrated satisfactory discriminatory power for predicting HF risk in T2DM patients, its measurement remains inaccessible in most primary healthcare settings in China. This study aimed to develop and externally validate a machine learning-based nomogram for predicting the risk of elevated NT-proBNP (≥125 pg/mL) as a surrogate for HF risk in patients with T2DM.
Methods: We retrospectively enrolled 564 T2DM patients as the development cohort and 302 from two external centers as the validation cohort. After feature selection via least absolute shrinkage and selection operator regression, five machine learning models were constructed and evaluated using 10-fold cross-validation. The optimal model was presented as a static nomogram and further deployed as an online web application for clinical use.
Results: Six key predictors were identified: estimated glomerular filtration rate, age, serum albumin, hemoglobin, urine albumin-to-creatinine ratio, and the binary indicator of age ≥ 65 years. Interpretability analysis using SHapley Additive exPlanations revealed estimated glomerular filtration rate as the most influential feature. The final machine learning-based nomogram achieved AUCs of 0.806 (95% CI: 0.767-0.845) in training and 0.861 (95% CI: 0.813-0.908) in external validation, with good calibration and clinical utility. Furthermore, the nomogram scores showed a significant positive correlation with established TRS-HFDM risk strata, supporting its clinical relevance.
Conclusion: We developed and validated an interpretable machine learning-based nomogram that effectively predicts the risk of elevated NT-proBNP in T2DM patients using six routine clinical variables. This tool demonstrates robust performance and generalizability, offering a practical and accessible solution for HF risk stratification in resource-limited primary care settings in China.

Keywords:  N-terminal pro-B-type natriuretic peptide; heart failure; machine learning; prediction model; type 2 diabetes mellitus

DOI:  https://doi.org/10.2147/DMSO.S558687
BMJ Open. 2025 Nov 21. 15(11): e101284

Aetiological clustering of newly diagnosed type 2 diabetes using machine learning: a retrospective cross-sectional study in Dubai, UAE.

Stafny Melony Dsouza, Fatima Sulaiman, Fatima Abdul, Fahad Mulla, Fayha Salah Ahmed, Mouza AlSharhan, Ayesha AlOlama, Noorah Ali, Aaesha Abdulaziz, Alia Mohammad Rafie, Suhaila Alnuaimi, Nandu Goswami, Amar Hassan Khamis, Riad Abdel Latif Bayoumi.

   OBJECTIVES: Type 2 diabetes (T2D) is a complex disease with a heterogeneous clinical presentation. Recently, five distinct clusters of T2D have been identified in the Emirati population of long-standing T2D with complications. This study aimed to validate these clusters in newly diagnosed T2D patients without any complications and determine whether severe and mild phenotypes are detectable early in the disease course.
DESIGN: Retrospective, cross-sectional, non-interventional study.
SETTING: Primary healthcare centres in Dubai, UAE.
PARTICIPANTS: A total of 451 adults, including both Emiratis and expatriates, diagnosed with T2D in the last 5 years and without T2D-related complications at the time of visit, were enrolled. Patients with complications, incomplete clinical data or higher duration of T2D were excluded from the study.
OUTCOME MEASURES: Identification of distinct T2D clusters using machine learning-based clustering analysis. Five clinical variables: age at diagnosis, body mass index, glycated haemoglobin, fasting serum insulin and fasting blood glucose served as predictors. Overlap between clusters was assessed via the Silhouette Index and Bayesian probability.
RESULTS: Five clusters were identified, replicating prior findings: severe insulin-resistant diabetes (SIRD), severe insulin-deficient diabetes (SIDD), mild age-related diabetes (MARD), mild obesity-related diabetes (MOD) and mild early-onset diabetes (MEOD). As confirmed by a Silhouette Index and Bayesian probability of 1, 55.43% of the patients showed cluster-exclusiveness, while 44.56% of the cohort showed overlap between clusters. The highest overlap was recorded for mild forms of T2D in the order MOD>MARD>MEOD.
CONCLUSIONS: The study confirms that both severe and mild T2D phenotypes are present in newly diagnosed, complication-free patients, supporting the applicability of cluster-based classification early in disease. These results highlight the potential for personalised treatment strategies to optimise management and prevent complications. Future studies should investigate longitudinal outcomes and therapeutic response across clusters.

Keywords:  Artificial Intelligence; Clinical chemistry; Diabetes Mellitus, Type 2; General diabetes; Machine Learning

DOI:  https://doi.org/10.1136/bmjopen-2025-101284
Chin J Nat Med. 2025 Nov;pii: S1875-5364(25)60969-1. [Epub ahead of print]23(11): 1301-1309

DeepGCGR: an interpretable two-layer deep learning model for the discovery of GCGR-activating compounds.

Xinyu Tang, Hongguo Chen, Guiyang Zhang, Huan Li, Danni Zhao, Zenghao Bi, Peng Wang, Jingwei Zhou, Shilin Chen, Zhaotong Cong, Wei Chen.

  The glucagon receptor (GCGR) is a critical target for the treatment of metabolic disorders such as Type 2 Diabetes Mellitus (T2DM) and obesity. Activation of GCGR enhances systemic insulin sensitivity through paracrine stimulation of insulin secretion, presenting a promising avenue for treatment. However, the discovery of effective GCGR agonists remains a challenging and resource-intensive process, often requiring time-consuming wet-lab experiments to synthesize and screen potential compounds. Recent advances in artificial intelligence technologies have demonstrated great potential in accelerating drug discovery by streamlining screening and efficiently predicting bioactivity. In the present work, we propose DeepGCGR, a two-layer deep learning model that leverages graph convolutional networks (GCN) integrated with a multiple attention mechanism to expedite the identification of GCGR agonists. In the first layer, the model predicts the bioactivity of various compounds against GCGR, efficiently filtering large chemical libraries to identify promising candidates. In the second layer, DeepGCGR classifies high bioactive compounds based on their functional effects on GCGR signaling, identifying those with potential agonistic or antagonistic effects. Moreover, DeepGCGR was specifically applied to identify novel GCGR-regulating compounds for the treatment of T2DM from natural products derived from traditional Chinese medicine (TCM). The proposed method will not only offer an effective strategy for discovering GCGR-targeting compounds with functional activation properties but also provide new insights into the development of T2DM therapeutics.

Keywords:  Artificial intelligence; Deep learning; G protein-coupled receptor; Natural products; Type 2 diabetes mellitus

DOI:  https://doi.org/10.1016/S1875-5364(25)60969-1
Diabetologia. 2025 Nov 17.

Artificial intelligence-driven clinical decision support systems to assist healthcare professionals and people with diabetes in Europe at the point of care: a Delphi-based consensus roadmap.

Mia Bajramagic, Tadej Battelino, Xavier Cos, Mark Cote, Nancy Cui, Angus Forbes, Alfonso Galderisi, Lutz Heinemann, Sufyan Hussain, Jessica Imbert, Christian Holm Jönsson, Michael Joubert, Nebojša M Lalić, Moshe Phillip, Peter Schwarz, Bart Torbeyns, Deborah J Wake, Katerina Zakrzewska, Stefano Del Prato.

  The use of artificial intelligence (AI) to improve the diagnosis, assessment and treatment of people with diabetes has the potential to drive a paradigm shift in diabetes care, both minimising treatment inertia and optimising clinical outcomes. This is a significant opportunity, given the predicted increase in the burden of diabetes over the next 20 years. However, there are concerns that regulatory processes for development and implementation of AI-driven technologies are not adequate for systems that may adapt to new data and change from their original performance characteristics as evaluated. The European Diabetes Forum (EUDF) convened a working group to review and investigate the unmet needs around implementation of AI technology in diabetes care. The working group developed the framework and focus of the accompanying analysis through a series of virtual and face-to-face meetings, including email conversations. The working group examined the key objectives for good diabetes care in the context of current and predicted AI-driven clinical decision support systems (AI-CDSS), including the outcomes for people with diabetes, the goals for personalised medicine and the implications for guideline-driven diabetes services and healthcare professionals. The process covered the needs of primary care healthcare professionals, who will shoulder the majority of diabetes care. The challenge of developing regulatory concepts and processes that are sufficiently robust to be AI inclusive was considered as central to the outcomes. Based on the available evidence, the EUDF working group believes that AI-CDSS will deliver benefits for people with diabetes, although there are clear challenges to moving AI-CDSS into the practical clinical space. To encourage debate on how this can be achieved safely and effectively, at the conclusion of the process a series of 14 recommendations was agreed using a nominal group technique and Delphi methodology, which are discussed in context in this article.

Keywords:  Artificial intelligence; Clinical decision support; Diabetes clinical practice; European Union; Regulatory process

DOI:  https://doi.org/10.1007/s00125-025-06601-5
Sci Rep. 2025 Nov 17. 15(1): 40228

My diabetes care: an AI-based mobile app with conversational agent for type 2 diabetes self-management.

T Ummal Sariba Begum, R Renuga Devi, Divya Haridas, Nebojsa Bacanin, Milica Djuric Jovicic, Bosko Nikolic.

  Despite advancements in modern healthcare, diabetes mellitus remains a lifelong, incurable condition. Empowering patients through health education and self-management is essential in preventing disease progression. This study evaluates the effectiveness of My Diabetes Care, a mobile application featuring an animated conversational agent, Dia-vera, designed to support diabetes self-managementat home. Focusing on non-compliance behaviors, sedentary lifestyle, and uncontrolled HbA1c levels, data were collected from 200 purposively selected participants from rural health clinics in southern Pakistan. This study used artificial intelligence models with built-in explainability features applied to artificial neural networks, achieving 98% training accuracy and 95% testing accuracy. User-chatbot dialogues were analyzed for engagement, thematic queries, fallback responses, and silence periods. Dia-vera successfully answered 88.86% of the 2830 queries. Weekly dialogue averages dropped from 36 to 26.1 between study phases, providing insights for future refinement. High levels of participant acceptability and satisfaction were found using the System Usability Scale. The findings show that, especially in disadvantaged settings, integrating interpretable AI with conversational agents provides a user-friendly and scientifically supported method of diabetes self-managementassistance.In comparison to baseline, participants who used the intervention reported better adherence to medication and food regimens, showed increased involvement in physical activity, and showed small reductions in HbA1c levels. These results make the study's therapeutic relevance stronger and show a stronger connection between the intervention and the desired health behaviors. Using My Diabetes Care as a proof-of-concept implementation, this study offers a reproducible framework for creating intelligent, explainable digital health interventions.

Keywords:  Artificial intelligence; Chat bot product development; Dia-Vera; Diabetes mellitus; Explainable artificial intelligence; Prediction; Self-management

DOI:  https://doi.org/10.1038/s41598-025-24172-8
Front Public Health. 2025 ;13 1708967

Patterns of discussion on neuroticism and self-management behaviors in type 2 diabetes: a scoping review using machine learning-assisted text mining.

Wei Sun, Dong Xie, Menglin Zhang, Wenhui Hou, Ziru Wang, Chendi Wang, Guodong Xu, He Yang.

   Background: Self-management behaviors, including diet control, medication adherence, blood glucose monitoring, and physical activity, are crucial for type 2 diabetes management. Neuroticism, a personality trait associated with anxiety and stress sensitivity, may significantly influence these behaviors. However, a comprehensive synthesis of evidence is lacking.
Objective: This scoping review aims to systematically map and synthesize how neuroticism has been examined in relation to self-management behaviors among adults with type 2 diabetes, and to identify recurring thematic patterns and knowledge gaps through machine learning-assisted text mining.
Methods: A scoping review was conducted in PubMed, Scopus, Web of Science, Embase, CINAHL, PsycINFO, and the Cochrane Library, covering the period from database inception to September 2025. The search strategy included keywords such as "neuroticism," "personality traits," "type 2 diabetes," "self-management," and "adherence." We used machine learning-assisted literature mining to summarize thematic patterns across included studies. The study selection process and workflow were conducted in accordance with the PRISMA-ScR guidelines.
Results: Ten studies were included. Across the literature, neuroticism was most frequently discussed alongside blood glucose monitoring, followed by diet control, medication taking, and exercise. Psychological constructs such as anxiety, stress sensitivity, and social support were commonly co-mentioned in these discussions. Machine learning-assisted analyses highlighted recurring topics, concept clusters, and co-occurrence patterns that characterize the discourse on neuroticism and T2DM self-management.
Conclusion: This scoping review characterizes how neuroticism is positioned within the discourse on T2DM self-management behaviors and delineates prominent thematic linkages and gaps. Machine learning-assisted text mining proved useful for organizing and visualizing dispersed evidence. Findings describe patterns in the literature rather than estimating causal effects, and can inform future hypothesis-driven studies and tailored clinical inquiry.
Systematic review registration: Unique Identifier: 10.17605/OSF.IO/54NJD; publicly accessible URL: https://doi.org/10.17605/OSF.IO/54NJD.

Keywords:  machine learning; neuroticism; self-management behaviors; text mining; type 2 diabetes mellitus

DOI:  https://doi.org/10.3389/fpubh.2025.1708967