bims-aukdir 2026-10-18 papers

bims-aukdir

Biomed News

on Automated knowledge discovery in diabetes research

Issue of 2026–10–18
fourteen papers selected by
Mott Given

Harnessing Artificial Intelligence-driven Approaches in the Diagnosis and Management of Diabetes and Diabetes-Associated Complications.
Research Progress on Artificial Intelligence in the Management of Diabetic Foot Ulcers.
Interpretable machine learning models to predict venous thromboembolism in sepsis patients with type 2 diabetes mellitus.
Automated detection of inflammation in foot thermographic images of patients with diabetes acquired in an uncontrolled clinical environment.
AI-Enhanced Wearable Glucose Sensors: From Intelligent Sensing to Closed-Loop Management of Diabetes.
AI-Assisted Nutrition Estimation and Exercise Support for Integrated Lifestyle Management in Patients with Diabetes.
Texture-aware diffusion modelling for severity-conditioned synthetic progression-like visualization of diabetic foot ulcers from single images.
Artificial intelligence-supported diabetic retinopathy screening in Tanzania: a randomised controlled trial.
Early Sound-Based Meal Onset Detection Using Hidden Markov Models (HMMs).
Cross-modal fusion of OCT and OCT angiography enface for improved diagnostics of diabetic retinopathy.
Machine learning models for identifying prediabetes: a systematic review of performance, validation, and implications for primary care.
Study on the Prediction of Diabetes-Cancer Comorbidity Based on Dietary Factors Explained by Seven Machine Learning Algorithms and SHapley Additive exPlanations.
Deep Learning-Derived Retinal Vascular Morphology and Incident Treatment for Sight-Threatening Diabetic Retinopathy.
AI-Driven Lifestyle Interventions for Reducing Diabetes Risk: Systematic Review.

Curr Diabetes Rev. 2026 Oct 05.

Harnessing Artificial Intelligence-driven Approaches in the Diagnosis and Management of Diabetes and Diabetes-Associated Complications.

Yachana Mishra, Rajeev Kanday, Simran Kauts, Sheetu Wadhwa, Aditi Chattaraj, Alaa A A Aljabali, Altijana Hromić-Jahjefendić, Vladimir N Uversky, Vijay Mishra.

  Diabetes mellitus is a global health issue with escalating metabolic dysfunction, clinical heterogeneity, and rising microvascular and macrovascular consequences. Traditional diabetes diagnosis, risk assessment, treatment selection, and complication surveillance rely on periodic clinical measures and population-level risk models, which may not fully capture dynamic, individualized disease pathways. Precision diabetes treatment is being transformed by Artificial intelligence (AI), including machine learning (ML), deep learning (DL), and multimodal data analytics. This review critically examines AI-driven approaches to diabetes management, including early disease prediction, diagnostic support, glycemic monitoring, treatment-response prediction, and the prevention and detection of diabetes-associated complications. The US Food and Drug Administration (USFDA) has sanctioned multiple AI/ML-based medical devices for automated retinal screening, clinical diagnostic assistance, and patient self-management applications. For clinically meaningful patterns and disease progression prediction, AI-enabled analysis of continuous glucose monitoring data, retinal images, electronic health records, wearable device signals, laboratory parameters, and multi-omics datasets is prioritized. The use of AI for detecting diabetic retinopathy, kidney disease, cardiovascular disease, neuropathy, and foot disorders is also highlighted.Data quality, algorithmic bias, interpretability, privacy, cybersecurity, model drift, regulatory monitoring, clinical validation, and equitable access prevent widespread application despite promising results. This study offers a thorough and critical assessment of the developing roles of AI, specifically ML and DL, in the prediction, diagnosis, and management of diabetes and its related complications. Real-world validation, standardized evaluation frameworks, trustworthy AI, and effective human- AI collaboration will drive advancement. AI could transform diabetes management into predictive, tailored, continuous, and complication-preventive care.

Keywords:  Artificial intelligence; deep learning; diabetes; diabetic retinopathy; glucose monitoring; insulin; machine learning

DOI:  https://doi.org/10.2174/0115733998484763260928100503
Diabetes Metab Syndr Obes. 2026 ;19 637013

Research Progress on Artificial Intelligence in the Management of Diabetic Foot Ulcers.

Jinxia Du, Xiaoli Liu, Guihui Wu, Yin Huang, Yan Zhang, Xiang Long, Xiaoyi Yang, Yan Zhou.

  As a typical chronic complication of diabetes, the management of diabetic foot ulcers has long faced multiple challenges, including the highly subjective nature of wound assessment, patients'limited capacity for self-care at home, and the difficulty in predicting healing outcomes. The emergence of artificial intelligence (AI) technology offers new approaches to addressing these challenges. Based on recent domestic and international literature, this paper systematically reviews the research progress of artificial intelligence in the care of diabetic foot ulcers, focusing on four key dimensions: intelligent wound assessment (eg deep learning-based image analysis models achieving an F1 score of 93.1% for infection identification and 92.8% accuracy for area assessment), patient self-health management (eg mobile health applications and wearable monitoring devices), the development and application of smart dressings (eg responsive biomaterials and sensor integration), and the prediction of healing trends (eg machine learning prediction models, with support vector machines achieving 85.3% accuracy and an AUC of 0.922 in predicting complete DFU healing). It summarizes key technological achievements and the current status of clinical translation, provides an in-depth analysis of existing shortcomings (including insufficient algorithm generalizability, lack of prospective clinical validation, and absence of data standardization), and outlines future development directions (encompassing multimodal integration, explainable artificial intelligence, and implementation science research), with the aim of offering guidance for the further application of artificial intelligence technology in the field of chronic wound care.

Keywords:  artificial intelligence; diabetic foot ulcers; review; smart dressings; wound care

DOI:  https://doi.org/10.2147/DMSO.S637013
Front Endocrinol (Lausanne). 2026 ;17 1958305

Interpretable machine learning models to predict venous thromboembolism in sepsis patients with type 2 diabetes mellitus.

Longyang Ma, Xiaoyuan Yu, Qi Xin.

   Background: Sepsis patients with type 2 diabetes mellitus (T2DM) face a substantially elevated risk of venous thromboembolism (VTE), yet prediction tools tailored for this population are lacking. This study aimed to develop and validate interpretable machine learning models for predicting in-hospital VTE in septic patients with T2DM.
Methods: A total of 648 sepsis patients with T2DM admitted to Shaanxi Provincial People's Hospital between January 2020 and October 2024 were included in this retrospective cohort study, an internal validation cohort of 303 patients was retrospectively collected from Shaanxi Provincial People's Hospital between November 2024 and December 2025, and an independent external validation cohort of 260 patients was retrospectively collected from Xi'an No. 3 Hospital (a separate tertiary care center) between November 2023 and December 2025. Six machine learning algorithms-Logistic Regression, Support Vector Machine (SVM), Multilayer Perceptron (MLP), LightGBM, XGBoost, and Random Forest-were developed and compared. Feature selection was performed using the Boruta algorithm combined with forward sequential feature selection. Model performance was evaluated by discrimination (area under the receiver operating characteristic curve, AUC), calibration (calibratin plots and Brier scores), and clinical utility (decision curve analysis). Model interpretability was enhanced using SHAP analysis.
Results: Among the 648 internal patients, 115 (17.7%) developed VTE. The Boruta algorithm and forward sequential feature selection identified five key predictors: D-dimer, C-reactive protein (CRP), activated partial thromboplastin time (APTT), international normalized ratio (INR), and fibrin degradation products (FDP). Among the six algorithms, LightGBM achieved the best overall performance, with an AUC of 0.9025 in the test set, 0.8048 in the internal validation cohort, and 0.8902 in the external validation cohort. Calibration curves demonstrated good agreement between predicted and observed outcomes, and decision curve analysis confirmed positive net clinical benefit across a wide range of threshold probabilities. SHAP analysis revealed that elevated D-dimer, CRP, INR, and FDP, together with shortened APTT, were the strongest drivers of VTE risk. Sensitivity analyses excluding D-dimer-driven cases, removing the D-dimer feature, and restricting to symptomatic VTE yielded consistent results (AUC range: 0.84-0.89).
Conclusions: The LightGBM model incorporating five readily available coagulation and inflammatory markers accurately predicts in-hospital VTE in sepsis patients with T2DM, with robust discrimination, calibration, and clinical utility. SHAP-based interpretability enhances transparency and may facilitate personalized risk stratification and informed thromboprophylaxis decisions in this high-risk population.

Keywords:  SHAP; machine learning; sepsis; type 2 diabetes mellitus; venous thromboembolism

DOI:  https://doi.org/10.3389/fendo.2026.1958305
Diabetes Res Clin Pract. 2026 Oct 12. pii: S0168-8227(26)00527-9. [Epub ahead of print]242 113607

Automated detection of inflammation in foot thermographic images of patients with diabetes acquired in an uncontrolled clinical environment.

Riki Wartakusumah, Hiroshi Noguchi, Saldy Yusuf, Kanae Mukai, Mayumi Okuwa, Makoto Oe.

   AIMS: Quantitative methods for detecting diabetic foot inflammation remain limited, though localized skin-temperature elevation is a promising early indicator. Because background variability can hinder AI-based foot identification, we evaluated AI-driven thermal imaging for diabetic foot inflammation under uncontrolled clinical backgrounds and assessed whether curtain-based background control and grayscale conversion improved detection.
METHODS: Cross-sectional study design was conducted in Makassar, Indonesia, 100 plantar thermal images from 50 diabetic patients were acquired with and without a curtain, preprocessed (sensitivity, 1.5; emissivity, 0.98), and converted to grayscale. Expert-identified inflammatory hotspots were annotated and used to train a YOLO11n model with 5-fold cross-validation, evaluated using accuracy, precision, sensitivity, specificity, ROC-AUC, IoU, and case-error analysis.
RESULTS: With a curtain, color-map images achieved sensitivity of 0.773, versus 0.455 without one. Grayscale images performed consistently across conditions, with best results in the mixed dataset (accuracy, 0.980; precision, 0.976; sensitivity, 0.932; specificity, 0.994; AUC 0.966 vs. 0.886 for color-map). Correctly detected cases showed IoU > 0.9. Low thermal contrast and small region of interest were the most frequent failure categories.
CONCLUSIONS: AI-driven thermal imaging detects diabetic foot inflammation under uncontrolled conditions. A curtain improved color-map detection, but grayscale imaging provided robust, curtain-independent performance, supporting its use for clinical screening.

Keywords:  Clinical settings; Grayscale image representation; Plantar thermography; Region detection method

DOI:  https://doi.org/10.1016/j.diabres.2026.113607
ACS Sens. 2026 Oct 12.

AI-Enhanced Wearable Glucose Sensors: From Intelligent Sensing to Closed-Loop Management of Diabetes.

Bo Sun, Lim Lee Ling, Mohamad Sofian Bin Abu Talip, Shixun Zheng, Yifu Fu, Quanjin Ma, Fazliyatul Azwa Md Rezali, Cheng Cheng, Yilin Peng, Sharifah F Wan Muhamad Hatta.

  Wearable glucose sensors enable continuous monitoring of glycemic dynamics, yet their clinical utility is constrained by signal instability, physiological delay, and limited predictive capability. Artificial intelligence (AI) has been increasingly integrated to address these limitations. This review summarizes recent advances in AI-enhanced wearable glucose sensors, with a focus on the evolution of sensing technologies and the role of AI across three main aspects: calibration, glucose prediction, and wearable closed-loop control. In calibration, AI-based approaches for improving short-term signal reliability and compensating long-term drift are reviewed. In prediction, methods ranging from continuous glucose monitoring (CGM)-based forecasting to multimodal sensor fusion and noninvasive sensing are summarized, including both continuous glucose estimation and event-oriented detection of hypo- and hyperglycemia. The integration of sensing, prediction, and insulin delivery in wearable closed-loop artificial pancreas is also discussed. Finally, the remaining challenges and future directions are discussed, highlighting the potential of AI-enhanced wearable glucose sensors to support more reliable and integrated diabetes management.

Keywords:  artificial intelligence; closed-loop glucose control; diabetes management; glycemic-event prediction; multimodal sensor fusion; wearable continuous glucose monitoring

DOI:  https://doi.org/10.1021/acssensors.6c02471
Nutrients. 2026 Sep 28. pii: 3207. [Epub ahead of print]18(19):

AI-Assisted Nutrition Estimation and Exercise Support for Integrated Lifestyle Management in Patients with Diabetes.

Muhammad Jamil, Adnan Kavak, Hossein Fotouhi, Md Rashed, Emre Gezer, Alpaslan Burak İnner, Gautam Srivastava.

  Background/Objectives: Multidimensional lifestyle interventions that combine healthy diet, physical activity, and psychosocial support are central to the effective self-management of type 1 and type 2 diabetes. Although digital health tools can improve lifestyle monitoring, existing AI-based dietary assessment methods often struggle to estimate food quantity accurately because the food size and scale are difficult to determine from images. This can lead to inconsistent estimates of food mass and energy content. This paper aimed to develop and technically evaluate PC2FoodNet, an AI-based framework for food recognition and physics-constrained physical quantity estimation, and to integrate dietary assessment and exercise support within the proposed AI-assisted Diabetes Care (AIDCare) mHealth platform for multidisciplinary lifestyle support. Methods: The AIDCare platform incorporates AI-assisted nutrition and professionally guided exercise modules to support personalized lifestyle management for individuals with diabetes. For dietary assessment, PC2FoodNet uses an EfficientNetV2-S backbone with multi-task regression and food-density priors to jointly model food volume, mass, and energy. A nutrition professional-in-the-loop approach supports the development and review of individualized diet plans and the assessment of patients' daily key nutrient intake according to individual nutritional requirements. The exercise module includes professionally designed exercises using Unreal Engine's MetaHuman plugin. The model was evaluated using four-fold stratified cross-validation on a dataset of 22,070 images covering 40 Turkish food categories. For physical quantity modeling, volume, mass, and energy targets were derived from standardized category-level references anchored to a 100 g reference portion rather than from image-specific ground-truth measurements. Results: PC2FoodNet achieved a mean Top-1 classification accuracy of 94.39% ± 0.47% and Top-5 accuracy of 98.90% ± 0.18% across the four folds. The corresponding Macro-F1 and Weighted-F1 scores were 93.14% ± 0.59% and 94.16% ± 0.36%, respectively. For the standardized category-level reference targets, the physics-constrained regression framework achieved mean MAE and RMSE values of 2.14 mL and 3.01 mL for reference volume, 21.65 g and 30.17 g for reference mass, and 41.20 kcal and 57.55 kcal for reference energy across all 22,070 pooled out-of-fold predictions. The confidence-based referral mechanism identified uncertain cases for clinical review, resulting in a clinical referral rate of 23.81%. Conclusions: AIDCare combines physics-constrained AI-based dietary assessment with personalized physical activity and continuous psychosocial support. A nutrition professional remains involved when AI predictions are uncertain, reducing the risks associated with fully automated lifestyle recommendations. The proposed approach provides a practical foundation for safer, more personalized, and evidence-based diabetes lifestyle management.

Keywords:  AIDCare; diabetes self-management; dietary assessment; food image recognition; mHealth; nutrition estimation; physical activity

DOI:  https://doi.org/10.3390/nu18193207
PLoS One. 2026 ;21(10): e0359540

Texture-aware diffusion modelling for severity-conditioned synthetic progression-like visualization of diabetic foot ulcers from single images.

Pratham Kaushik, Vinay Kukreja, Eshika Jain, Modafar Ati, Shanmugasundaram Hariharan.

Diabetic foot ulcers (DFUs) are serious complications of diabetes associated with delayed healing, infection, and limb amputation. Modelling DFU progression remains challenging because publicly available image datasets generally contain independent single-timepoint images rather than longitudinal follow-up sequences. This study proposes a texture-aware conditional diffusion framework for generating severity-conditioned, progression-like synthetic DFU images from a single RGB input. Instead of claiming patient-specific temporal prediction, the framework models plausible population-level visual transitions across ordinal pseudo-stages derived from clinically relevant severity cues. The full model achieved an SSIM of 0.851 ± 0.020, PSNR of 24.76 ± 1.12 dB, and FID of 44.33 ± 5.02, with higher observed SSIM and PSNR and lower FID values than the evaluated GAN and VAE baselines across structural fidelity and distributional realism. The GAN baseline achieved SSIM 0.782 ± 0.031, PSNR 21.98 ± 1.56 dB, and FID 59.13 ± 6.59, while the VAE baseline achieved SSIM 0.738 ± 0.038, PSNR 19.63 ± 1.83 dB, and FID 69.77 ± 7.44. The framework also obtained a Texture Consistency Score of 0.79 ± 0.04 and a Granulation Tissue Fidelity Score of 0.76 ± 0.05, indicating preservation of fine-grained texture and granulation-related appearance patterns. Ablation results confirmed the contribution of texture conditioning and cross-attention to image fidelity. Advanced severity-state analysis produced SSIM values of 0.674, 0.804, and 0.726 across three pseudo-stages, supporting severity-aligned structural consistency. Since true longitudinal future-state images were unavailable, these metrics indicate synthetic image fidelity and distributional realism, not longitudinal progression accuracy. Longitudinal DFU cohorts and larger clinician-centred validation remain necessary before clinical decision-support use or temporal progression claims can be justified. To assess generalization beyond the primary 493-image dataset, external validation was additionally performed on an independent DFU dataset under a zero-shot setting, demonstrating stable cross-dataset performance with only modest degradation in image fidelity and texture consistency. Therefore, the proposed framework should be interpreted as modelling cross-sectional severity-conditioned appearance interpolation across independent cases, rather than true patient-specific future progression prediction.

DOI: https://doi.org/10.1371/journal.pone.0359540
Lancet Glob Health. 2026 Oct 12. pii: S2214-109X(26)00224-X. [Epub ahead of print] 104062

Artificial intelligence-supported diabetic retinopathy screening in Tanzania: a randomised controlled trial.

Charles R Cleland, Covadonga Bascaran, William U Makupa, Bernadetha Shilio, Justus Rwiza, Lusekelo Nyale, Edith Macha, Fatima Mumba, Franki Sandi, Heiko Philippin, Ana P Marques, Louis Bolter, John Anderson, Catherine Egan, Adnan Tufail, Pearse A Keane, Alastair K Denniston, David Macleod, Matthew J Burton.

BACKGROUND: Non-attendance after referral from screening is one of the major challenges for diabetic retinopathy screening programmes in Africa. Artificial intelligence (AI) systems for the identification of diabetic retinopathy can provide point-of-screening results and therefore enable immediate counselling. We aimed to investigate whether AI-supported diabetic retinopathy screening, with a point-of-screening result and face-to-face counselling, increased the proportion of people with true referable diabetic retinopathy who attended the central ophthalmology referral clinic when compared with standard diabetic retinopathy screening with deferred human grading and telephone counselling.
METHODS: This two-arm, parallel-group, individually randomised controlled trial was conducted at non-communicable disease or diabetic clinics in 24 hospitals in the Kilimanjaro and Arusha regions of Tanzania. Eligible participants (aged ≥18 years) had diabetes and attended diabetic retinopathy screening. Participants were randomly assigned (1:1) to receive either AI-supported diabetic retinopathy screening with a point-of-screening result and face-to-face counselling (the AI-intervention group) or standard-of-care diabetic retinopathy screening with deferred human retinal image assessment and telephone counselling (the standard-of-care group). Participants had retinal photographs taken of both eyes. Images from the AI-intervention group were submitted to the AI software, and these participants were given their screening result (non-referable, referable, or ungradable) plus face-to-face counselling before leaving the photography room, with a referral to the central ophthalmology clinic if the AI output was referable or ungradable. Participants in the standard-of-care group received their screening result, along with appropriate counselling, by telephone a few weeks later. The primary outcome was attendance at the central ophthalmology clinic within 8 weeks of screening among participants classified by masked UK-certified graders as having true referable diabetic retinopathy. This trial was registered with the ISRCTN Registry, ISRCTN18317152, and is complete.
FINDINGS: Between March 8, 2023, and Aug 31, 2024, 2439 participants were assessed for eligibility and 2378 were enrolled, of whom 1192 were randomly assigned to the AI-intervention group and 1186 to the standard-of-care group. 904 participants had true referable diabetic retinopathy, comprising 437 in the AI-intervention group and 467 in the standard-of-care group. Of the participants in the AI-intervention group, 157 (35·9%) of 437 attended the central ophthalmology clinic within 8 weeks, compared with 85 (18·2%) of 467 in the standard-of-care group, an absolute difference of 17·7 percentage points (95% CI 12·0-23·4; odds ratio 2·52 [95% CI 1·86-3·42]; p<0·0001).
INTERPRETATION: AI-supported screening, with a point-of-screening result and face-to-face counselling, increased the proportion of people with true referable diabetic retinopathy that attended the central ophthalmology clinic after screening compared with the standard of care in Tanzania. Such AI-enabled screening pathways have the potential to improve diabetic retinopathy screening programmes in Africa.
FUNDING: British Council for the Prevention of Blindness, Christian Blind Mission, and the Sir Halley Stewart Trust.

DOI: https://doi.org/10.1016/j.langlo.2026.104062
IEEE J Transl Eng Health Med. 2026 ;14 410-424

Early Sound-Based Meal Onset Detection Using Hidden Markov Models (HMMs).

Laura Lema-Perez, Maja Grauff Erntsen, Oyvind Stavdahl, Anders Lyngvi Fougner.

   OBJECTIVE: Meal detection is essential for automated insulin delivery systems in patients with Type 1 Diabetes (T1D), as meals are the primary drivers of rapid postprandial glucose excursions. Current closed-loop systems depend on manual meal announcements, which are often inaccurate or omitted, reducing safety and autonomy. To address this, we developed an automatic meal detector leveraging physiological sound sequences from the body.
METHODS AND PROCEDURES: The dataset comprised 50 meal recordings from ten participants (aged 24-30 years) using a throat-mounted microphone. Twenty recordings were collected in quiet conditions, while 30 included deliberate, realistic disturbances. Audio signals were preprocessed, segmented, and transformed into 46 features. Linear Discriminant Analysis (LDA) was implemented as a benchmark, while continuous Hidden Markov Model (HMM)s were applied to capture temporal dependencies and estimate meal onset probability.
RESULTS: Performance was evaluated using both an 80:20 train-test split and Leave-One-Subject-Out Cross-Validation (LOSO CV). Explainable Artificial Intelligence (XAI) techniques were employed to interpret model behavior and identify informative features of meal events. LDA achieved 90% accuracy, 78% sensitivity, and a detection delay of 12.3 s, while HMM reached 91% for both accuracy and sensitivity, with an average detection time of 11.7 s. Both models performed well under quiet conditions but were sensitive to background disturbances.
CONCLUSION: These findings demonstrate the feasibility of detecting meal onset from physiological sounds. XAI analysis revealed that HMM effectively captures sequential transitions associated with chewing and swallowing, offering interpretable insights beyond non-temporal methods. Thus, HMM shows potential as a rapid, context-aware meal classifier for closed-loop insulin therapy, thanks to its effective modeling of temporal structure.
CLINICAL IMPACT: This pilot study uses an innovative throat-mounted microphone to detect early meal intake by capturing chewing and swallowing sounds, offering an alternative to traditional measures such as continuous glucose monitoring (CGM).

Keywords:  Continuous hidden Markov model; diabetes; explainable AI; feature extraction; hyperparameter optimization; machine learning

DOI:  https://doi.org/10.1109/JTEHM.2026.3732514
Front Med (Lausanne). 2026 ;13 1900570

Cross-modal fusion of OCT and OCT angiography enface for improved diagnostics of diabetic retinopathy.

Rashadul Hasan Badhon, Atalie Carina Thompson, Jennifer I Lim, Theodore Leng, Minhaj Nur Alam.

   Introduction: Diabetic retinopathy (DR) is a leading cause of vision impairment worldwide, highlighting the need for accurate and accessible screening tools. Optical Coherence Tomography (OCT) provides high-resolution structural information of the retina, whereas OCT angiography (OCTA) offers complementary vascular information that is highly relevant for DR diagnosis.
Methods: In this study, we propose a cross-modal fusion of OCT B-scans with single-channel enface OCTA using a bidirectional cross-modal attention network for automated DR classification. Two independent datasets, OCT500 and UIC, comprising 730 scans in total, were utilized to evaluate performance in within-dataset, combined-dataset, and cross-dataset generalization settings. A ConvNeXt V2 Large model trained solely on OCT images served as the unimodal baseline. In addition to ground-truth (GT) OCTA, we explored the use of translated (TR) OCTA generated from OCT scans, eliminating the requirement for dedicated OCTA hardware.
Results: Experimental results demonstrated that cross-modal fusion consistently outperformed unimodal OCT classification across all evaluation scenarios. Fusion with GT OCTA improved classification accuracy and discriminative performance, while TR OCTA achieved comparable or superior results in most settings. Furthermore, TR OCTA improved sensitivity and cross-dataset generalization, indicating enhanced robustness to domain shifts.
Discussion: These findings demonstrate that attention-based OCTOCTA enface fusion provides clinically meaningful improvements for DR detection and suggest that computationally generated OCTA can serve as a practical, low-cost alternative to hardware-acquired OCTA, enabling broader deployment of high-performance retinal screening systems in resource-limited clinical environments.

Keywords:  DR; OCT; OCTA; OCTA enface; classification; crossmodal fusion

DOI:  https://doi.org/10.3389/fmed.2026.1900570
Front Med (Lausanne). 2026 ;13 1956800

Machine learning models for identifying prediabetes: a systematic review of performance, validation, and implications for primary care.

Liang Wang, Yan Zhang, Hao Yang, Xiwei Wang, Xinyuan Qi, Li Gou.

   Background: Machine learning models have been proposed for identifying prediabetes, but their methodological quality, generalisability, and readiness for primary care implementation remain uncertain.
Methods: We conducted a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020-compliant systematic review registered in PROSPERO (CRD420251029165). Eight databases were searched from inception to 1 April 2025, with PubMed, Embase, and Web of Science Core Collection updated through 1 August 2026. The primary review question focused on current/prevalent prediabetes identification; longitudinal/prognostic studies addressing future dysglycaemia were retained only as supplementary evidence and interpreted separately. Principal performance and implementation conclusions were restricted to current/prevalent identification. Two reviewers independently screened studies and extracted data. Risk of bias and applicability were assessed using Prediction Model Risk of Bias Assessment Tool (PROBAST). Findings were synthesised descriptively because of heterogeneity.
Results: Twenty-six studies involving 435,647 participants were included. Of these, 23 studies constituted the primary evidence base for current/prevalent prediabetes identification, whereas three longitudinal/prognostic studies were retained as supplementary evidence and interpreted separately. Across all included evidence, 110 models were reported. Development or internal validation area under the receiver operating characteristic curves (AUCs) ranged from 0.49 to 0.9989 across heterogeneous modelling tasks; the highest value represented overall multiclass discrimination rather than a prediabetes-specific binary AUC. Independent external validation was reported in only one study (1/26, 3.8%), in which the area under the receiver operating characteristic curve (AUROC) for prediabetes decreased from 0.746 in internal evaluation to 0.590 and 0.523 in two external cohorts. Fourteen studies were judged to have a low overall risk of bias, whereas 12 studies (46.2%) were judged to have a high overall risk of bias. Calibration, clinical utility assessment, and subgroup fairness evaluation remained limited.
Conclusion: Machine learning models show potential to support current/prevalent prediabetes identification in primary care, particularly when based on routinely collected variables. However, strong performance observed during model development or internal validation should not be interpreted as evidence of implementation readiness. Given the limited independent external validation and insufficient evidence regarding calibration, clinical utility, fairness, and real-world workflow integration, current evidence remains insufficient to support routine primary care implementation. These conclusions relate to current/prevalent prediabetes identification and should not be extrapolated to longitudinal/prognostic prediction of future prediabetes, impaired fasting glucose, or related dysglycaemic progression.
Systematic Review Registration: PROSPERO CRD420251029165.

Keywords:  machine learning; prediabetes; prediction model; primary care; screening; systematic review

DOI:  https://doi.org/10.3389/fmed.2026.1956800
J R Soc N Z. 2026 Oct;56(5): e70088

Study on the Prediction of Diabetes-Cancer Comorbidity Based on Dietary Factors Explained by Seven Machine Learning Algorithms and SHapley Additive exPlanations.

Youcheng Yang, Bin Fang.

  Background: The bidirectional link between diabetes and cancer is well established, but the role of dietary factors in predicting their comorbidity remains unclear. This study aims to develop a machine learning model to integrate dietary components and to identify key predictors using SHAP analysis. Methods: Data were derived from the NHANES 2009-2012 and 2015-2018 cycles, including 20,307 participants (682 with diabetes-cancer comorbidity). After eliminating collinearity and standardizing, 41 dietary features were retained. Seven algorithms-Naive Bayes, support vector machine with radial basis kernel and class weights (SVM-RW), random forest (RF), weighted k-nearest neighbors (weighted k-NN), AdaBoost, LogitBoost, and cancerclass-were trained and evaluated by benchmark tests using AUC-ROC and AUC-PR. The best-performing model was interpreted with SHAP values. Results: AdaBoost achieved the highest discrimination, with AUC-ROC of 0.898 and AUC-PR of 0.851. SHAP analysis revealed that vitamins (D, K, A, C, thiamin, total folate, niacin, and β-carotene), minerals (calcium, iron, sodium, and copper), and carbohydrates/dietary fiber were the most influential features. Among these, carbohydrates and calcium exerted the dominant negative contribution to comorbidity prediction, indicating that higher intake was associated with a lower predicted probability of diabetes-cancer comorbidity by the model. Conclusions: The AdaBoost model exhibited the best performance for predicting diabetes-cancer comorbidity. Dietary components played an important role in disease prediction, among which carbohydrates and calcium served as core negative predictive factors. The identified associations provide clues and evidence for further research on dietary interventions for high-risk populations.

Keywords:  SHapley Additive exPlanation; cancer; diabetes; diet; machine learning

DOI:  https://doi.org/10.1002/snz2.70088
Ophthalmol Sci. 2026 Nov;6(11): 101385

Deep Learning-Derived Retinal Vascular Morphology and Incident Treatment for Sight-Threatening Diabetic Retinopathy.

Paul Nderitu, Mark Chia, Urvi Karamchandani, Christine A Kiire, Asterios Diafas, Sarah Matta, Yukun Zhou, Pearse A Keane.

   Objective: Retinal vascular morphology (RVM) features could be useful biomarkers for risk stratification for incident sight-threatening diabetic retinopathy (STDR). We evaluated whether deep learning-derived macula-centered RVM, extracted at scale, are associated with incident treatment for STDR including proliferative diabetic retinopathy (PDR) and clinically significant macula edema (CSME).
Design: Retrospective longitudinal cohort study (2000-2022).
Participants: We included 20 855 patients (33 861 eyes) with gradable, treatment-naïve eyes with a macula-centered fundus image at baseline.
Methods: Arterial and venous RVM features of complexity, density, caliber, and tortuosity were extracted using an automated deep learning tool (AutoMorph). Multivariable Cox proportional hazards models were performed to quantify the association between RVM and incident treatment for STDR, adjusting for age, sex, ethnicity, baseline diabetic retinopathy (DR) and diabetic macula edema (DME) grade, and intereye correlation using a cluster term.
Main Outcome Measures: Incident treatment for STDR including PDR or CSME.
Results: Over a mean 1.74 (standard deviation 2.15) years of follow-up, 3446 eyes (10.2%) received treatment for STDR (1387 PDR; 2059 CSME). Higher arterial complexity and density were strongly protective for incident STDR treatment (arterial complexity Q4 vs. Q1 hazard ratio [HR] 0.52 [95% confidence interval 0.45-0.60]; arterial density HR 0.47 [0.41-0.54]). Larger venous caliber was associated with an increased risk of incident STDR treatment (HR 2.53 [2.23-2.86]), while increasing arterial caliber was associated with a reduced risk of incident PDR treatment (HR 0.64 [0.53-0.77]). There were weak associations between increasing arterial and venous tortuosity and an increased risk of incident treatment for PDR and CSME, respectively (arterial tortuosity HR 1.11 [95% confidence interval 1.01-1.22]; venous tortuosity HR 1.16 [1.03-1.31]). Retinal vascular morphology feature associations were generally stronger for incident treatment for PDR than CSME. Associations between RVM and incident treatment for STDR remained consistent after a sensitivity analysis which only included patients with no or mild DR and no DME at baseline.
Conclusions: Macula-centered RVM features show clinically plausible associations with incident treatment for STDR. Retinal vascular morphology features extracted using an automated deep learning tool are potential prognostic biomarkers that could be useful for DR risk prediction models alongside other retinal imaging biomarkers and clinical risk factors.
Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Keywords:  AutoMorph; Deep learning; Diabetic retinopathy; Prognostic biomarkers; Retinal vascular morphology

DOI:  https://doi.org/10.1016/j.xops.2026.101385
J Med Internet Res. 2026 Oct 15. 28 e104940

AI-Driven Lifestyle Interventions for Reducing Diabetes Risk: Systematic Review.

Obinna I Ekwunife, Tyler Nauta, Raphael A Fraser, Jennifer A Campbell, Gloria Aidoo-Frimpong, Rebekah J Walker, Leonard E Egede.

   BACKGROUND: Artificial intelligence (AI)-driven lifestyle interventions are increasingly used for reducing diabetes risk. However, whether they produce clinical benefit beyond standard Diabetes Prevention Programs (DPPs) and whether effects vary by AI type, population, or delivery setting, remains uncertain.
OBJECTIVE: This study aimed to evaluate the effects of AI-driven lifestyle interventions on clinical, behavioral, and engagement outcomes among people with prediabetes or elevated diabetes risk and to assess whether effects differed by AI type, population, or delivery setting.
METHODS: PubMed, Embase, CINAHL, Web of Science, IEEE Xplore, ClinicalTrials.gov, and citation searches were completed on August 3, 2026, for English-language reports published from 2015. Eligible studies evaluated AI-driven digital lifestyle interventions among people with prediabetes or elevated diabetes risks, or healthy participants in feasibility studies. One reviewer (OIE) screened records and extracted data; a second reviewer (TN) independently verified exclusions and extraction. Study quality was assessed with National Heart, Lung, and Blood Institute tools appropriate to each design. Given substantial heterogeneity in designs, estimands, and outcome metrics, findings were synthesized narratively following the Synthesis Without Meta-Analysis guideline rather than pooled statistically.
RESULTS: Of 8032 records, 11 studies were included: 1 randomized, 1 nonrandomized comparative, 4 observational, and 5 uncontrolled pre-post studies. Seven evaluated AI lifestyle coaching systems, 2 predictive AI platforms, and 2 conversational AI chatbots. Ten of 11 studies used commercially developed platforms. Study quality was rated good for 5 studies, fair for 3, and poor for 3. The randomized controlled trial showed noninferiority of an AI-DPP versus human coach-led DPP for the combined 12-month outcome (23/183, 12.6% vs 23/185, 12.4%; risk difference 0.1%, 1-sided 95% CI -5.5%). Program initiation (171/183, 93.4% vs 153/185, 82.7%; risk difference 10.7%, 95% CI 4.2%-17.3%) and completion (117/183, 63.9% vs 93/185, 50.3%; risk difference 13.7%; 95% CI 3.6%-23.7%) were higher with AI, but this did not translate into superior clinical outcomes. The nonrandomized comparison found no advantage of AI-personalized messaging over static messaging in terms of BMI or physical activity. The remaining studies reported uncontrolled changes precluding causal inference. Evidence was insufficient to determine whether effects differed by population or delivery setting.
CONCLUSIONS: One randomized trial supports noninferiority, not superiority, of a specific AI-driven DPP relative to human coaching; the remaining evidence was nonrandomized, predominantly commercially developed, and hypothesis-generating. To our knowledge, this is the first systematic review to stratify evidence on AI-driven diabetes prevention by AI mechanism-coaching, predictive, and conversational-and by outcome domain, extending prior scoping work that mapped apps without appraising study quality or comparative effectiveness. This structured mapping shows where evidence is the strongest (engagement and weight) and the weakest (glycemic causality and generalizability), giving health systems a concrete basis for deciding where AI-DPP tools may supplement, rather than replace, standard DPP delivery, pending independent, adequately powered randomized trials.

Keywords:  artificial intelligence; diabetes prevention; digital health; lifestyle interventions; prediabetes; systematic review

DOI:  https://doi.org/10.2196/104940