bims-aukdir Biomed News
on Automated knowledge discovery in diabetes research
Issue of 2026–01–18
thirteen papers selected by
Mott Given
  1. Predicting the diabetic foot in patients with type 2 diabetes mellitus based on machine learning.
  2. Pediatric diabetes prediction using machine learning.
  3. Assessment of Correlation Between Diabetic Retinopathy and Metabolic Biomarkers Using Artificial Intelligence.
  4. Diagnostic Accuracy of the EyeArt Artificial Intelligence System for Diabetic Retinopathy: A Systematic Review and Meta-Analysis.
  5. Diabetic retinopathy detection and grading system using deep learning approach.
  6. Systematic Review on Deep Learning Algorithms for Blood Glucose Forecasting in Type 1 Diabetes.
  7. SDC-Net: Enhancing Multiclass Diabetic Retinopathy Segmentation by Shallow-Deep Collaboration Networks With Wavelet-Guided Attention Mechanism.
  8. AI-Enhanced Imaging for Diabetic Foot Ulcer Risk Assessment and Diagnosis: A Retrospective Cohort Study.
  9. Predicting postpartum glucose intolerance in women with gestational diabetes mellitus in primary care: A machine learning approach using XGBoost and SHAP values.
  10. A foundation model for continuous glucose monitoring data.
  11. Automated method for quantitative analysis of iris fluorescein angiography based on machine learning.
  12. Precision phenotyping of type 2 diabetes in chinese populations using a variational autoencoder-informed tree model.
  13. Multiomics Integration Analysis Reveals the Regulatory Mechanisms of Efferocytosis in Diabetic Kidney Disease.
  1. Biomed Eng Online. 2026 Jan 16.
    Predicting the diabetic foot in patients with type 2 diabetes mellitus based on machine learning.
    Haixiang Zhang, Weijian Fan, Peipei Li, Xiangzi Chen, Shiwu Yin.
       BACKGROUND: Diabetic foot (DF) is a severe complication of type 2 diabetes mellitus (T2DM), contributing to significant morbidity and healthcare costs globally. Early prediction and intervention are critical for preventing amputations and improving patient outcomes. However, traditional statistical methods lack the capacity to handle high-dimensional clinical data and identify optimal predictive features. This study aimed to develop and validate machine learning models for DF risk prediction using feature selection strategies based on binary logistic regression and information theory.
    METHODS: A retrospective cohort of 1,179 patients (95 DF cases, 1,084 T2DM controls) was analyzed using clinical and biochemical data from 2019 to 2025. Three data sets were constructed: (1) original features; (2) features selected via binary logistic regression (F1); and (3) features selected via information-theoretic global learning (F2). Six models-extreme learning machine (ELM), kernel extreme learning machine (KELM), and their variants trained on the three data sets-were evaluated using fivefold cross-validation. Performance metrics included area under the receiver operating characteristic curve (AUC), sensitivity, specificity, and computational efficiency.
    RESULTS: Age, blood-urea-nitrogen (BUN), homocysteine (Hcy), albumin (ALB), and fasting blood glucose (FBG) were identified as independent DF risk factors. The information theory-based KELM (IT-KELM) model achieved the highest AUC of 0.799 (sensitivity: 0.792 and specificity: 0.710) on F2, outperforming other models. Feature selection improved predictive accuracy while reducing computational time, with IT-KELM requiring 0.138 s for training and 0.0023 s for testing. The SHAP summary dot plot and bar chart revealed that the top five features contributing to the model were TP, RBC, ALB, BMI and HB.
    CONCLUSIONS: Integrating information theory with KELM enhances DF risk prediction by optimizing feature subsets and leveraging nonlinear kernel mapping. The IT-KELM model demonstrates robust diagnostic performance and clinical feasibility for early DF screening. Future multi-center studies are needed to validate generalizability and refine model interpretability in real-world settings. This approach provides a cost-effective tool for precision medicine in diabetes care.
    Keywords:  Diabetic foot; Extreme learning machine; Feature selection; Information theory; Machine learning; Type 2 diabetes mellitus
    DOI:  https://doi.org/10.1186/s12938-025-01494-2
  2. Sci Rep. 2026 Jan 15. 16(1): 1979
    Pediatric diabetes prediction using machine learning.
    Abeer El-Sayyid El-Bashbishy, Hazem M El-Bakry.
      Diabetes is a chronic condition that affects a substantial portion of the global population and is linked to elevated mortality rates and a range of severe health complications. Despite its clinical importance, progress in diabetes research is often constrained by the limited availability of comprehensive datasets and robust predictive models. To address these challenges, researchers are increasingly turning to big data analytics and machine learning (ML) methodologies. This study presents the development of an ML-based system aimed at predicting the likelihood of diabetes and classifying its various types. A novel dataset, termed Diabetes Types Dataset, was constructed by integrating four heterogeneous dataset sources: paediatrics data from the Mansoura University Children Hospital repository, the Pima Indian Diabetes (PIMA) dataset, the Pone dataset, and a Gestational Diabetes dataset. The classification of diabetes types was approached as a multiclass problem using a suite of supervised ML algorithms, including Artificial Neural Networks (ANN), Logistic Regression, Naive Bayes, Decision Trees, Adaptive Boosting, Random Forests, Gradient Boosting, Support Vector Machines, and K-Nearest Neighbors. Model performance was evaluated using several metrics: Accuracy, Precision, Mean Squared Error, and Area Under the Receiver Operating Characteristic Curve. Among the models tested, the ANN classifier demonstrated the highest accuracy, achieving a peak performance of 99.98%. Further validation was conducted using an external dataset referred to as diabetes_prediction, which confirmed the model's robustness with consistent accuracy. Additionally, the proposed system was applied to a publicly available dataset, diabetes_Dataset, containing 34 features used to predict 12 distinct types of diabetes efficiently. The results suggest that this ML-driven approach can significantly enhance the ability of healthcare professionals to detect and classify diabetes types, thereby supporting early intervention and improved disease management.
    DOI:  https://doi.org/10.1038/s41598-025-24964-y
  3. J Diabetes Res. 2026 ;2026 9085827
    Assessment of Correlation Between Diabetic Retinopathy and Metabolic Biomarkers Using Artificial Intelligence.
    Mustafa Aydemir, Ahmet Burak Bilgin, Ramazan Sari, Mehmet Erkan Doğan, Mehmet Bulut, Yusuf Akar.
       Background: One of the main causes of blindness in the world, diabetic retinopathy (DR) is a dangerous condition that impairs vision in diabetics. Preventing visual loss requires early recognition of DR and prompt treatments. Artificial intelligence (AI) software combined with nonmydriatic fundus cameras has demonstrated encouraging gains in DR screening effectiveness. However, there are not many studies that systematically compare the diagnostic effectiveness of various nonmydriatic cameras and AI software in the field of endocrinology, where managing diabetes and its complications is crucial. By offering vital information for enhancing diabetes care plans and fortifying preventative actions in the context of endocrine health, this study seeks to close this knowledge gap.
    Methods: This clinical study was conducted at the Akdeniz University endocrinology clinic with 900 volunteer patients who had previously been diagnosed with diabetes but had undiagnosed DR. Fundus images of each patient were captured using three different nonmydriatic fundus cameras. These images were then assessed for varying degrees of DR, ranging from mild to more severe forms, including vtDR and clinically significant diabetic macular edema, utilizing EyeCheckup AI software. Additionally, patients underwent pupil dilation for wide-angle fundus photography, resulting in four distinct wide-angle images being taken. Three retina specialists evaluated these four wide-field fundus images based on the DR treatment guidelines set forth by the American Academy of Ophthalmology. The effectiveness of the AI in detecting DR was determined through statistical analysis, comparing the diagnoses made by the physicians with those provided by the AI. Furthermore, patients filled out a questionnaire regarding their medical history and underwent a lipid panel blood test along with urine tests. These assessments included various metabolic measurements such as HbA1c levels, diabetes duration, low-density lipoprotein (LDL) cholesterol, high-density lipoprotein (HDL) cholesterol, triglycerides, urinary albumin levels, glomerular filtration rate (GFR), creatinine, and C-reactive protein (CRP) levels.
    Results: Our study revealed a significant association between the prevalence of DR and diabetes duration, HbA1c, CRP, and urinary albumin levels. The p values of this association were 0.000, 0.000, 0.003, and 0.002, respectively. It was also noted that there may be an association between triglyceride levels and DR prevalence; the p value of this association was 0.079, with more data needed to establish a strong link. The study also revealed that AI performed satisfactorily in detecting DR from fundus images. The sensitivity and specificity of the different cameras used are as follows: Canon CR2 AF: 95.65% sensitivity, 95.92% specificity; Topcon TRC-NW400: 95.19% sensitivity, 96.46% specificity; and Optomed Aurora: 90.48% sensitivity, 97.21% specificity.
    Conclusion: Our research has shown a significant association between the prevalence of DR and elevated levels of HbA1c, CRP, and urinary albumin and duration of diabetes. This suggests that these biomarkers may serve as valuable predictive indicators in assessing the likelihood of DR. Consequently, the inclusion of these parameters in routine clinical assessments could improve proactive screening strategies, thus enabling early detection and intervention of DR. This, in turn, could reduce the risk of vision loss in affected patients. The study also demonstrates the potential of nonmydriatic fundus cameras used in combination with AI software to detect DR at an early stage.
    Trial Registration: ClinicalTrials.gov identifier: NCT04805541.
    Keywords:  artificial intelligence; diabetes mellitus; diabetic retinopathy screening; metabolic biomarkers
    DOI:  https://doi.org/10.1155/jdr/9085827
  4. Ophthalmologica. 2026 Jan 13. 1-29
    Diagnostic Accuracy of the EyeArt Artificial Intelligence System for Diabetic Retinopathy: A Systematic Review and Meta-Analysis.
    Isabel Inmaculada Guedes Guedes, Pedro Saavedra Santana, Ángel Ramos de Miguel, Ángel Ramos Macías, Francisco Cabrera López, Ayoze González Hernández.
       INTRODUCTION: Diabetic retinopathy (DR) persists as a predominant cause of preventable vision loss globally, with its prevalence escalating in conjunction with the diabetes epidemic. Efficient, automated screening is needed to enable earlier detection of DR at scale. Artificial intelligence (AI)-driven platforms, such as EyeArt® (Eyenuk Inc.), offer a scalable solution with potential to alleviate the burden on healthcare systems.
    METHODS: A systematic review (SR) and meta-analysis were conducted following PRISMA and MOOSE guidelines. This review was prospectively registered in PROSPERO (CRD42024571137). Observational studies published between 2016 and 2024 assessing the diagnostic performance of the EyeArt® system for DR detection were retrieved from PubMed, Scopus, and Embase. Data on sensitivity, specificity, and diagnostic odds ratio (DOR) were extracted, and pooled estimates were calculated using a random-effects model. Study quality was assessed using QUADAS-2 and GRADE frameworks.
    RESULTS: Seventeen studies, met the inclusion criteria. The pooled log diagnostic odds ratio (LDOR) was 3.96 (95% CI 3.54-4.39), and the area under the summary receiver operating characteristic (SROC) curve was 0.932 (95% CI 0.885-0.985), indicating high overall diagnostic accuracy. No significant heterogeneity was observed in the pooled diagnostic OR, although sensitivity and specificity varied across studies.
    CONCLUSIONS: EyeArt® demonstrates high diagnostic accuracy for detecting any-grade and referable DR across diverse clinical and geographical settings. Its integration into DR screening programs could improve early detection, optimize healthcare resource allocation, and expand access to ophthalmic care, particularly in resource-limited environments.
    KEY MESSAGES: • EyeArt®demonstrated high diagnostic accuracy for detecting referable or any-grade DR across diverse settings. • Its consistent performance supports its integration into routine DR screening workflows. • Deployment of EyeArt®for DR may optimize resource allocation, streamline diagnostic pathways, and expand access, particularly in resource-limited environments.
    DOI:  https://doi.org/10.1159/000550443
  5. Digit Health. 2026 Jan-Dec;12:12 20552076251410982
    Diabetic retinopathy detection and grading system using deep learning approach.
    D Siva Sundhara Raja, S Kumarganesh, K Martin Sagayam, Hien Dang.
      Early screening for diabetic retinopathy (DR) prevents vision loss in diabetic patients.
    Objective: To classify DR cases, this study suggests computer-assisted screening and diagnosis. The proposed methodology consists of retinal blood vessel, macular region, and exudate segmentation. In this study, exudates were screened, and DR was classified as mild, moderate, or severe by analyzing the presence of exudates in the macular region.
    Methods: The bit-plane morphological slicing technique was employed to segment the macular region. The U-Net deep learning approach was used to segment the exudates from the retinal images, and the proposed methodologies were applied and evaluated on retinal images in the publicly available Methods to Evaluate Segmentation and Indexing Techniques in the field of Retinal Ophthalmology (MESSIDOR) and High-Resolution Fundus (HRF) datasets. Finally, DR can be diagnosed by analyzing the presence of exudates in the macular region, and DR can be diagnosed as mild, moderate, or severe based on the severity levels. The conventional segmentation algorithms segment and locate the internal boundary of pixels in both exudates and macula. In contrast with the conventional segmentation methods, the proposed algorithm segments and locates both internal and external boundary of pixels in both exudates and macula region which increases the segmentation accuracy.
    Results: The proposed macular region segmentation method obtained 98.9% SeI, 99.2% SpI, and 99.1% AccI for the MESSIDOR dataset. The proposed macular region segmentation method obtained 99.2% SeI, 99.7% SpI, and 99.8% AccI for the HRF dataset. The proposed exudate segmentation method obtained 99.3% SeI, 99.2% SpI, and 99.2% AccI for the MESSIDOR dataset. The proposed exudate segmentation method obtained 99.7% SeI, 99.3% SpI, and 99.1% AccI for the HRF dataset.
    Conclusion: The segmentation time period of both exudates and macula region in retinal image is low when comparing with other segmentation methods. Moreover, this proposed segmentation and diagnosis method was tested with other DIARETDB1, Retinal Fundus Multi-disease Image, and Fundus Image Registration Dataset datasets to validate the effectiveness of the proposed method.
    Keywords:  Diabetic retinopathy; exudates; macula region; vision loss
    DOI:  https://doi.org/10.1177/20552076251410982
  6. IEEE J Biomed Health Inform. 2026 Jan 14. PP
    Systematic Review on Deep Learning Algorithms for Blood Glucose Forecasting in Type 1 Diabetes.
    Andrea Calzavara, Francesco Prendin, Giacomo Cappon, Simone Del Favero, Andrea Facchinetti.
      Type 1 Diabetes (T1D) is a chronic metabolic disease characterized by elevated blood glucose (BG) concentrations, resulting from the immune-mediated destruction of insulin-producing $\beta$-cells in the pancreas. Effective management of T1D greatly benefits from constant monitoring of BG levels, achievable in real-time using minimally invasive continuous glucose monitoring (CGM) devices. These devices provide data streams that can be leveraged by forecasting algorithms to predict BG levels minutes in advance, enabling timely therapeutic interventions to prevent adverse events, such as hypo/hyperglycemia. With the increasing availability of data, deep learning (DL) algorithms have emerged as the state-of-the-art for BG forecasting, owing to their ability to autonomously learn complex nonlinear relationships, such as those underlying the glucoregulatory system. Despite a growing body of research, a comprehensive review specifically focusing on DL applications for BG prediction is still lacking. To address this gap, a systematic review was conducted following the PRISMA guidelines, involving extensive searches across PubMed, Scopus, and Web of Science databases. A total of 26 studies satisfied the inclusion criteria and were evaluated based on dataset characteristics, model inputs, training paradigm, prediction horizon, model architecture, evaluation metrics, performance, and baseline comparators. While DL models show great promise, several challenges persist-particularly in ensuring physiological fidelity and interpretability, both essential for clinical adoption. To overcome these barriers, future research should prioritize the integration of explainable AI (XAI) techniques to improve model reliability and safety, ultimately supporting the effective deployment of DL models in real-time T1D management.
    DOI:  https://doi.org/10.1109/JBHI.2025.3630214
  7. Transl Vis Sci Technol. 2025 Dec 01. 14(12): 28
    SDC-Net: Enhancing Multiclass Diabetic Retinopathy Segmentation by Shallow-Deep Collaboration Networks With Wavelet-Guided Attention Mechanism.
    Jiaoli Liu, Lin Zhang, Zhen Liu, Yonggang Xu, Xiaojuan Wang, Hui Liu.
       Purpose: Diabetic retinopathy (DR) is a leading cause of blindness. Fundus lesions are key clinical signs of DR, and their accurate segmentation is crucial for screening, grading, and monitoring the disease. However, segmenting different lesions simultaneously is challenging owing to their varying shapes, sizes, and appearances. This work aimed to segment four types of DR lesions simultaneously.
    Methods: We propose a shallow-deep collaboration network with a wavelet-guided attention mechanism for simultaneous segmentation of four DR-related lesions. Our end-to-end framework integrates shallow and deep networks to enhance multiscale feature extraction. The deep network uses a wavelet-based attention mechanism to fuse multiscale context representations. Additionally, a super-resolution auxiliary task is introduced to improve training accuracy.
    Results: Extensive experiments are conducted on the IDRiD, DDR, and FGADR public datasets for evaluation. The average Dice scores of SDC-Net on the three datasets are 60.27, 39.98, and 42.57, with average intersection over union scores of 44.53, 25.71, and 28.58, and average area under the receiver operating characteristic curve values of 68.75, 57.37, and 64.21.
    Conclusions: The dual-branch design in the proposed framework enables better capture of multiscale features and improves segmentation. The dual wavelet attention module of the deep branch can enhance the extraction of detailed lesion features, and the introduced super-resolution task further improves the accuracy and robustness.
    Translational Relevance: The SDC-Net framework has the potential to enhance the clinical diagnosis of DR by offering more precise segmentation of multiple types of fundus lesions, thereby aiding in early detection and management of the disease.
    DOI:  https://doi.org/10.1167/tvst.14.12.28
  8. J Diabetes Sci Technol. 2026 Jan 15. 19322968251409761
    AI-Enhanced Imaging for Diabetic Foot Ulcer Risk Assessment and Diagnosis: A Retrospective Cohort Study.
    Tilak Bhattacharya, Sandip Chakraborty, Ghanshyam Goyal, Manisha Singh, B Edward Jude, Saswati Mukherjee.
       BACKGROUND: The automated assessment and prediction of diabetic foot ulcer (DFU) severity depends heavily on precise segmentation of the ulcer region. This approach avoided reliance on built-in segmentation tools, which often lacked the accuracy needed to delineate wound boundaries effectively. The objective of this study was to develop and evaluate an artificial intelligence (AI)-driven method for ulcer segmentation and severity classification of DFU using Wagner's grading system.
    METHODS: A novel method was introduced for segmenting the boundaries of DFUs, paired with a lightweight classification model for predicting ulcer severity as per Wagner's grade. This method was developed using a retrospective cohort of patients in India. A total of 1339 ulcer images were collected from 510 patients and augmented to 6579 images for AI-model generalizability. It incorporated an enhanced active contour model, combined with Sobel edge detection, to achieve precise delineation of ulcer edges. An AI-powered mobile application was developed to facilitate the real-time and remote assessment of the severity of DFUs.
    RESULTS: The proposed segmentation approach successfully delineated ulcer regions, achieving a Dice similarity coefficient of 0.99. The classification model attained an accuracy of 95.58%, with a sensitivity of 95.58%, a specificity of 99.16%, and an F1 score of 95.53%. The method also recorded a false-positive rate of 0.84% and a false negative rate of 4.83%, reflecting improved classification performance compared to existing methods.
    CONCLUSIONS: The comparative analysis demonstrated that the proposed method significantly improved both segmentation and classification of DFUs, thereby supporting enhanced clinical management of the condition.
    Keywords:  MobileNetV3-Small; Sobel operator; Wagner grading system; active contour model; diabetic foot ulcer; segmentation
    DOI:  https://doi.org/10.1177/19322968251409761
  9. Diabetes Res Clin Pract. 2026 Jan 12. pii: S0168-8227(26)00017-3. [Epub ahead of print] 113098
    Predicting postpartum glucose intolerance in women with gestational diabetes mellitus in primary care: A machine learning approach using XGBoost and SHAP values.
    Nur Aiza Idris, Nurul Azreen Yusof, Muhd Zulfadli Hafiz Ismail, Siti Norazlina Juhari, Nurulhuda Mat Hassan, Norwati Daud, Nurul Izza Yunus, Mohd Faeiz Pauzi.
       OBJECTIVE(S): To develop and internally validate an interpretable machine learning model using eXtreme Gradient Boosting (XGBoost) and Shapley Additive exPlanations (SHAP) to predict postpartum glucose intolerance among women with GDM using routine antenatal clinical data.
    STUDY DESIGN: This retrospective study included 600 women with GDM who completed a 6-week postpartum 75-g OGTT. Forty-three antenatal variables were extracted from electronic medical records. An XGBoost model was trained using stratified 5-fold cross-validation, ROSE oversampling, and grid-search optimisation. Model performance was evaluated using AUC, precision, recall, F1 score and negative predictive value (NPV). SHAP analysis was used to assess feature importance and interpretability.
    RESULTS: Postpartum glucose intolerance occurred in 19% of participants. The XGBoost model achieved an AUC of 0.671 and PR-AUC of 0.35, with precision of 0.79, recall of 0.82 and an NPV of 0.87. SHAP analysis identified fasting plasma glucose, 2-hour glucose, gestational weight gain, multiparity, previous GDM and family history of diabetes as key predictors.
    CONCLUSION: An interpretable XGBoost model with SHAP explanations using routine antenatal data shows promise for postpartum glucose risk assessment in primary care. Despite moderate predictive performance, the model demonstrated a high negative predictive value.
    Keywords:  Gestational diabetes; Glucose intolerance; Machine learning; Postpartum period; Primary health care
    DOI:  https://doi.org/10.1016/j.diabres.2026.113098
  10. Nature. 2026 Jan 14.
    A foundation model for continuous glucose monitoring data.
    Guy Lutsker, Gal Sapir, Smadar Shilo, Jordi Merino, Anastasia Godneva, Jerry R Greenfield, Dorit Samocha-Bonet, Raja Dhir, Francisco Gude, Shie Mannor, Eli Meirom, Eric P Xing, Gal Chechik, Hagai Rossman, Eran Segal.
      Continuous glucose monitoring (CGM) generates detailed temporal profiles of glucose dynamics, but its full potential for achieving glucose homeostasis and predicting long-term outcomes remains underutilized. Here we present GluFormer, a generative foundation model for CGM data trained with self-supervised learning on more than 10 million glucose measurements from 10,812 adults mainly without diabetes1,2. Using autoregressive prediction, the model learned representations that transferred across 19 external cohorts (n = 6,044) spanning 5 countries, 8 CGM devices and diverse pathophysiological states, including prediabetes, type 1 and type 2 diabetes, gestational diabetes and obesity. These representations provided consistent improvements over baseline blood glucose and HbA1c levels and other CGM-derived measures for forecasting glycaemic parameters3,4. In individuals with prediabetes, GluFormer stratified those likely to experience clinically significant increases in HbA1c over a 2-year period, outperforming baseline HbA1c and common CGM metrics. In a cohort of 580 adults with short-term CGM and a median follow-up of 11 years5, GluFormer identified individuals at elevated risk of diabetes and cardiovascular mortality more effectively than HbA1c. Specifically, 66% of incident diabetes cases and 69% of cardiovascular deaths occurred in the top risk quartile, compared with 7% and 0%, respectively, in the bottom quartile. In clinical trials, baseline CGM representations improved outcome prediction. A multimodal extension of the model that integrates dietary data generated plausible glucose trajectories and predicted individual glycaemic responses to food. Together, these findings indicate that GluFormer provides a generalizable framework for encoding glycaemic patterns and may inform precision medicine approaches for metabolic health.
    DOI:  https://doi.org/10.1038/s41586-025-09925-9
  11. Quant Imaging Med Surg. 2026 Jan 01. 16(1): 68
    Automated method for quantitative analysis of iris fluorescein angiography based on machine learning.
    Yixuan Zhu, Shuo Sun, Shaolei Han, Jing Chen, David Western, Longli Zhang.
       Background: Diabetic retinopathy is a leading cause of vision impairment, often progressing to neovascular glaucoma. Early detection of neovascularisation of the iris (NVI) is crucial for timely intervention. Traditional diagnostic methods, such as slit-lamp examination, have limitations in identifying early-stage NVI. This study presents a deep learning-based automated approach for analysing iris fluorescein angiography (IFA) images to detect and quantify peripupillary leakage, a key indicator of NVI.
    Methods: A dataset of 2,449 IFA images was used to train a YOLOv8n-based segmentation model for precise pupil localisation. A leakage circularity detection algorithm was developed to quantify peripupillary fluorescein leakage. The algorithm's performance was evaluated using an independent test set of 131 clinically standardized IFA images. Performance metrics included mean absolute error (MAE), mean absolute percentage error (MAPE), and intersection over union (IoU). Results were compared with manual annotations from two clinical experts.
    Results: The proposed method demonstrated a significant reduction in MAE (20.81 degrees) and MAPE (21.64%) compared to Clinical Staff 1 (MAE: 34.23 degrees, MAPE: 58.38%) and Clinical Staff 2 (MAE: 43.17 degrees, MAPE: 75.71%). The algorithm achieved an IoU of 39.3%, slightly lower than Clinical Staff 1 (44.5%) and Clinical Staff 2 (41.7%), indicating high segmentation accuracy but minor spatial misalignment. The inter-clinician agreement yielded an IoU of 54.8%, highlighting subjectivity in human assessments.
    Conclusions: The deep learning-based approach provides superior consistency and accuracy in quantifying peripupillary fluorescein leakage compared to manual expert annotations. While human experts demonstrated slightly higher spatial precision, the algorithm significantly reduces variability and subjectivity in leakage quantification. This automated method has the potential to enhance early detection of NVI, improve clinical workflow efficiency, and assist ophthalmologists in diagnosing DR. Further optimization will focus on refining spatial segmentation accuracy.
    Keywords:  Deep learning; diabetic retinopathy (DR); iris fluorescein angiography (IFA); iris leakage detection; neovascular glaucoma
    DOI:  https://doi.org/10.21037/qims-2025-480
  12. Nat Commun. 2026 Jan 14.
    Precision phenotyping of type 2 diabetes in chinese populations using a variational autoencoder-informed tree model.
    Tong Yue, Wenhao Zhang, Yu Ding, Xueying Zheng, Yunjie Ma, Juliana C N Chan, Eric S H Lau, Juliana N M Lui, Guoxi Jin, Wen Xu, Yan Bi, Zuocheng Wang, Sheng Nie, Mengchun Gong, Ewan R Pearson, Sihui Luo, Jianping Weng.
      Type 2 diabetes (T2D) exhibits clinical heterogeneity, yet most existing classification models are derived from European populations and face challenges in clinical application. Here, we evaluate the generalizability of a tree-like graph structure from Scottish data to 32,501 newly diagnosed T2D patients from a multi-center Chinese cohort comprising over 8.6 million individuals. We observe similar distribution between the Scottish and Chinese individuals in heart and kidney outcomes, but diabetic retinopathy varies across ancestries even within similar phenotypes. To capture T2D Chinese-specific heterogeneity, we apply a variational autoencoder (VAE) framework to identify key clinical features and construct a tree structure using the Discriminative Dimensionality Reduction Tree (DDRTree) algorithm. This Chinese tree model is validated in two independent external cohorts and revealed longitudinal phenotypic shifts trending toward higher-risk branches. Our findings emphasize the need for population-specific classification frameworks to advance precision diabetology through individualized risk prediction and specialized treatment guidelines.
    DOI:  https://doi.org/10.1038/s41467-025-68211-4
  13. Inflammation. 2026 Jan 13.
    Multiomics Integration Analysis Reveals the Regulatory Mechanisms of Efferocytosis in Diabetic Kidney Disease.
    Yi Kang, Qian Jin, Mengqi Zhou, Huijuan Zheng, Aoshuang Li, Danwen Li, Xuezhe Wang, Jingwei Zhou, Jie Lv, Yaoxian Wang.
      Diabetic kidney disease (DKD) is a prevalent complication in individuals with diabetes. Efferocytosis plays a pivotal role in chronic diseases; however, the precise mechanisms involved in DKD are still not fully understood. DKD-related datasets were obtained from the Gene Expression Omnibus (GEO) database, and differentially expressed genes (DEGs) were screened. These DEGs subsequently intersected with efferocytosis-related genes (ERGs) to produce DKD efferocytosis‒related genes (DKD-ERGs). Potential hub genes were subsequently identified using protein‒protein interaction (PPI) network analysis in combination with machine learning (LASSO regression, Boruta algorithm, and random forest algorithm). Next, we employed transcriptomics, proteomics, and metabolomics analyses of DKD animal models, followed by validation with serum samples from patients with DKD. A nomogram was developed using hub genes to evaluate its predictive accuracy. Consensus clustering was utilized to categorize DKD patients and conduct immune infiltration analysis. A total of 15 DKD-related ERGs were identified. ANXA1, CASP3, IL33, and C3 were identified as potential hub genes. First, validation was performed using the GEO and Nephroseq databases. The hub genes were subsequently validated from multiple perspectives, including transcriptomics, metabolomics, and proteomics of DKD animal models, as well as serological analysis of DKD patients. A risk score model incorporating these 4 hub genes effectively predicted both the onset and progression of DKD. On the basis of these hub genes, DKD patients were classified into Cluster 1 and Cluster 2, with distinct subtypes and immune infiltration correlating with disease stages. This study reveals the potential diagnostic value of ERGs (ANXA1, CASP3, IL33, and C3) in DKD through multidimensional analysis. These genes may serve as promising biomarkers and therapeutic targets for DKD.
    Keywords:  Biomarkers; Diabetic kidney disease; Efferocytosis; Immune infiltration; Machine learning
    DOI:  https://doi.org/10.1007/s10753-025-02401-6
This page is being maintained by Thomas Krichel. It was last updated on 2026‒01‒18 at 0:10.