Supervised Parametric Learning in the Identification of Composite Biomarker Signatures of Type 1 Diabetes in Integrated Parallel Multi-Omics Datasets DOI Creative Commons
J. T. Bonnell, Óscar Garnica, B.M. Watts

et al.

Biomedicines, Journal Year: 2024, Volume and Issue: 12(3), P. 492 - 492

Published: Feb. 22, 2024

Background: Type 1 diabetes (T1D) is a devastating autoimmune disease, and its rising prevalence in the United States around world presents critical problem public health. While some treatment options exist for patients already diagnosed, individuals considered at risk developing T1D who are still early stages of their disease pathogenesis without symptoms have no any preventive intervention. This because uncertainty determining level predicting with high confidence will progress, or not, to clinical diagnosis. Biomarkers that assess one’s certainty could address this inform decisions on intervention, especially children where burden justifying high. Single omics approaches (e.g., genomics, proteomics, metabolomics, etc.) been applied identify biomarkers based specific disturbances association disease. However, reliable remained elusive date. To overcome this, we previously showed parallel multi-omics provides more comprehensive picture disease-associated facilitates identification candidate biomarkers. Methods: paper evaluated use machine learning (ML) using data augmentation supervised ML methods purpose improving salient patterns ultimate extraction novel biomarker candidates integrated datasets from limited number samples. We also examined different integration (early, intermediate, late) which stage parametric models can learn under conditions dimensionality variation feature counts across omics. In late scheme, employed multi-view ensemble comprising individual trained over single computational challenges posed by yet datasets. Results: improves prediction case vs. control finds most success flagging larger consistent set associated features when compared chance models, may eventually be used downstream identifying composite signature risk. Conclusions: current work demonstrates utility exploring ongoing quest biomarkers, reinforcing hope signatures via ultimately informing face escalating global incidence debilitating

Language: Английский

Artificial intelligence for diabetes care: current and future prospects DOI
Bin Sheng, Krithi Pushpanathan, Zhouyu Guan

et al.

The Lancet Diabetes & Endocrinology, Journal Year: 2024, Volume and Issue: 12(8), P. 569 - 595

Published: July 23, 2024

Language: Английский

Citations

29
AI-Driven Innovations in Alzheimer's Disease: Integrating Early Diagnosis, Personalized Treatment, and Prognostic Modelling DOI
Mayur B. Kale, Nitu L. Wankhede,

Rupali S. Pawar

et al.

Ageing Research Reviews, Journal Year: 2024, Volume and Issue: unknown, P. 102497 - 102497

Published: Sept. 1, 2024

Language: Английский

Citations

18
Advancements and applications of Artificial Intelligence in cardiology: Current trends and future prospects DOI Creative Commons
David B. Olawade, Nicholas Aderinto, Gbolahan Olatunji

et al.

Journal of Medicine Surgery and Public Health, Journal Year: 2024, Volume and Issue: 3, P. 100109 - 100109

Published: April 23, 2024

Using Artificial intelligence technologies in cardiology has witnessed rapid advancements across various domains, fostering innovation and reshaping clinical practices. The study aims to provide a comprehensive overview of these AI-driven their implications for enhancing cardiovascular healthcare. A systematic approach was adopted conduct an extensive review scholarly articles peer-reviewed literature focusing on the application AI cardiology. Databases including PubMed/MEDLINE, ScienceDirect, IEEE Xplore, Web Science were systematically searched. Articles screened following defined selection criteria. These articles' synthesis highlighted AI's diverse applications cardiology, but not limited diagnostic innovations, precision medicine, remote monitoring technologies, drug discovery, decision support systems. shows significant role medicine by revolutionising diagnostics, treatment strategies, patient care. showcased this reflect transformative potential technologies. However, challenges such as algorithm accuracy, interoperability, integration into workflows persist. continued strategic promise deliver more personalised, efficient, effective care, ultimately improving outcomes shaping future practice.

Language: Английский

Citations

12
AI-based diabetes care: risk prediction models and implementation concerns DOI Creative Commons
Serena Wang,

Grace Nickel,

Kaushik P. Venkatesh

et al.

npj Digital Medicine, Journal Year: 2024, Volume and Issue: 7(1)

Published: Feb. 15, 2024

The utilization of artificial intelligence (AI) in diabetes care has focused on early intervention and treatment management. Notably, usage expanded to predict an individual's risk for developing type 2 diabetes. A scoping review 40 studies by Mohsen et al. shows that while most used unimodal AI models, multimodal approaches were superior because they integrate multiple types data. However, creating models determining model performance are challenging tasks given the multi-factored nature For both there also concerns bias with lack external validations representation race, age, gender training barriers data quality evaluation standardization ripe areas new technologies, especially entrepreneurs innovators. Collaboration amongst providers, entrepreneurs, researchers must be prioritized ensure is providing equitable patient care.

Language: Английский

Citations

9
A novel interpretable deep transfer learning combining diverse learnable parameters for improved T2D prediction based on single-cell gene regulatory networks DOI Creative Commons

Sumaya Alghamdi,

Turki Turki

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Feb. 24, 2024

Abstract Accurate deep learning (DL) models to predict type 2 diabetes (T2D) are concerned not only with targeting the discrimination task but also useful feature representation. However, existing DL tools far from perfect and do provide appropriate interpretation as a guideline explain promote superior performance in target task. Therefore, we an interpretable approach for our presented transfer (DTL) overcome such drawbacks, working follows. We utilize several pre-trained including SEResNet152, SEResNeXT101. Then, knowledge via keeping weights convolutional base (i.e., extraction part) while modifying classification part use of Adam optimizer deal classifying healthy controls T2D based on single-cell gene regulatory network (SCGRN) images. Another DTL work similar manner just bottom layers unaltered updating consecutive through training scratch. Experimental results whole 224 SCGRN images using five-fold cross-validation show that model (TFeSEResNeXT101) achieving highest average balanced accuracy (BAC) 0.97 thereby significantly outperforming baseline resulted BAC 0.86. Moreover, simulation study demonstrated superiority is attributed distributional conformance weight parameters obtained when coupled model.

Language: Английский

Citations

5
Prediction model for type 2 diabetes mellitus and its association with mortality using machine learning in three independent cohorts from South Korea, Japan, and the UK: a model development and validation study DOI Creative Commons
Hayeon Lee, Soon Cheon Hwang, Seoyoung Park

et al.

EClinicalMedicine, Journal Year: 2025, Volume and Issue: 80, P. 103069 - 103069

Published: Jan. 18, 2025

Language: Английский

Citations

0
Unveiling diabetes onset: Optimized XGBoost with Bayesian optimization for enhanced prediction DOI Creative Commons
Muhammad Khurshid,

Sadaf Manzoor,

Touseef Sadiq

et al.

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(1), P. e0310218 - e0310218

Published: Jan. 24, 2025

Diabetes, a chronic condition affecting millions worldwide, necessitates early intervention to prevent severe complications. While accurately predicting diabetes onset or progression remains challenging due complex and imbalanced datasets, recent advancements in machine learning offer potential solutions. Traditional prediction models, often limited by default parameters, have been superseded more sophisticated approaches. Leveraging Bayesian optimization fine-tune XGBoost, researchers can harness the power of data analysis improve predictive accuracy. By identifying key factors influencing risk, personalized prevention strategies be developed, ultimately enhancing patient outcomes. Successful implementation requires meticulous management, stringent ethical considerations, seamless integration into healthcare systems. This study focused on optimizing hyperparameters an XGBoost ensemble model using optimization. Compared grid search (accuracy: 97.24%, F1-score: 95.72%, MCC: 81.02%), with achieved slightly improved performance 97.26%, MCC:81.18%). Although improvements observed this are modest, optimized represents promising step towards revolutionizing treatment. approach holds significant outcomes for individuals at risk developing diabetes.

Language: Английский

Citations

0
REMED-T2D: A robust ensemble learning model for early detection of type 2 diabetes using healthcare dataset DOI

Le Thi Phan,

Rajan Rakkiyappan,

Balachandran Manavalan

et al.

Computers in Biology and Medicine, Journal Year: 2025, Volume and Issue: 187, P. 109771 - 109771

Published: Feb. 5, 2025

Language: Английский

Citations

0
Development and validation of prediction models for stroke and myocardial infarction in type 2 diabetes based on health insurance claims: does machine learning outperform traditional regression approaches? DOI Creative Commons
Anna-Janina Stephan, Michael Hanselmann, Medina Bajramovic

et al.

Cardiovascular Diabetology, Journal Year: 2025, Volume and Issue: 24(1)

Published: Feb. 18, 2025

Abstract Background Digitalization and big health system data open new avenues for targeted prevention treatment strategies. We aimed to develop validate prediction models stroke myocardial infarction (MI) in patients with type 2 diabetes based on routinely collected high-dimensional insurance claims compared predictive performance of traditional regression state-of-the-art machine learning including deep methods. Methods used German from 2014 2019 287 potentially relevant literature-derived variables predict 3-year risk MI stroke. Following a train-test split approach, we the logistic methods without forward selection, LASSO-regularization, random forests (RF), gradient boosting (GB), multi-layer-perceptrons (MLP) feature-tokenizer transformers (FTT). assessed discrimination (Areas Under Precision-Recall Receiver-Operator Curves, AUPRC AUROC) calibration. Results Among n = 371,006 (mean age: 67.2 years), 3.5% ( 13,030) had MIs 3.4% 12,701) strokes. AUPRCs were 0.035 0.034 (stroke) null model, between 0.082 0.092 (GB) MI, 0.061 0.073 stoke. AUROCs 0.5 models, 0.70 (RF, MLP, FTT) 0.71 (all other models) 0.66 0.69 All well calibrated. Conclusions Discrimination claims-based reached ceiling at around 0.09 0.7 AUROC. While AUROC this was comparable existing epidemiological incorporating clinical information, comparison other, more metrics, such as AUPRC, sensitivity Positive Predictive Value hampered by lack reporting literature. The fact that did not outperform approaches may suggest feature richness complexity exploited before choice algorithm could become critical maximize performance. Future research might focus impact different derivation ceilings. In absence powerful screening alternatives, applying transparent regression-based routine claims, though certainly imperfect, remains promising scalable low-cost approach population-based cardiovascular stratification. Graphical abstract

Language: Английский

Citations

0
Reporting Quality of AI Intervention in Randomized Controlled Trials in Primary Care: Systematic Review and Meta-Epidemiological Study DOI Creative Commons
Jingru Zhong, Ting Zhu, Yafang Huang

et al.

Journal of Medical Internet Research, Journal Year: 2025, Volume and Issue: 27, P. e56774 - e56774

Published: Feb. 25, 2025

Background The surge in artificial intelligence (AI) interventions primary care trials lacks a study on reporting quality. Objective This aimed to systematically evaluate the quality of both published randomized controlled (RCTs) and protocols for RCTs that investigated AI care. Methods PubMed, Embase, Cochrane Library, MEDLINE, Web Science, CINAHL databases were searched until November 2024. Eligible studies or full exploring was assessed using CONSORT-AI (Consolidated Standards Reporting Trials–Artificial Intelligence) SPIRIT-AI (Standard Protocol Items: Recommendations Interventional checklists, focusing intervention–related items. Results A total 11,711 records identified. In total, 19 21 RCT 35 included. overall proportion adequately reported items 65% (172/266; 95% CI 59%-70%) 68% (214/315; 62%-73%) protocols, respectively. percentage specific item ranged from 11% (2/19) 100% (19/19) 10% (2/21) (21/21), exhibited similar characteristics trends. They lack transparency completeness, which can be summarized three aspects: without providing adequate information regarding input data, mentioning methods identifying analyzing performance errors, stating whether how intervention its code accessed. Conclusions could improved protocols. helps promote transparent complete with

Language: Английский

Citations

0