A systematic review of machine learning models for management, prediction and classification of ARDS

Respiratory Research, Journal Year: 2024, Volume and Issue: 25(1)

Published: June 4, 2024

Abstract Aim Acute respiratory distress syndrome or ARDS is an acute, severe form of failure characterised by poor oxygenation and bilateral pulmonary infiltrates. Advancements in signal processing machine learning have led to promising solutions for classification, event detection predictive models the management ARDS. Method In this review, we provide systematic description different studies application Machine Learning (ML) artificial intelligence management, prediction, classification We searched following databases: Google Scholar, PubMed, EBSCO from 2009 2023. A total 243 was screened, which, 52 were included review analysis. integrated knowledge previous work providing state art overview explainable decision identified areas future research. Results Gradient boosting most common successful method utilised 12 (23.1%) studies. Due limitation data size available, neural network its variation used only 8 (15.4%) Whilst all cross validating technique separated database validation, 1 study validated model with clinician input. Explainability methods presented 15 (28.8%) feature importance which 14 times. Conclusion For databases 5000 fewer samples, extreme gradient has highest probability success. large, multi-region, multi centre required reduce bias take advantage method. framework explaining ML clinicians involved would be very helpful development deployment model.

Language: Английский

---

21st century critical care medicine: An overview

World Journal of Critical Care Medicine, Journal Year: 2024, Volume and Issue: 13(1)

Published: March 5, 2024

Critical care medicine in the 21st century has witnessed remarkable advancements that have significantly improved patient outcomes intensive units (ICUs). This abstract provides a concise summary of latest developments critical care, highlighting key areas innovation. Recent include Precision Medicine: Tailoring treatments based on individual characteristics, genomics, and biomarkers to enhance effectiveness therapies. The objective is describe recent Care Medicine. Telemedicine: integration telehealth technologies for remote monitoring consultation, facilitating timely interventions. Artificial intelligence (AI): AI-driven tools early disease detection, predictive analytics, treatment optimization, enhancing clinical decision-making. Organ Support: Advanced life support systems, such as Extracorporeal Membrane Oxygenation Continuous Renal Replacement Therapy provide better organ support. Infection Control: Innovative infection control measures combat emerging pathogens reduce healthcare-associated infections. Ventilation Strategies: ventilation modes lung-protective strategies minimize ventilator-induced lung injury. Sepsis Management: Early recognition aggressive management sepsis with tailored Patient-Centered Care: A shift towards patient-centered focusing psychological emotional well-being addition medical needs. We conducted thorough literature search PubMed, EMBASE, Scopus using our strategy, incorporating keywords telemedicine, management. total 125 articles meeting criteria were included qualitative synthesis. To ensure reliability, we focused only published English language within last two decades, excluding animal studies,

Language: Английский

---

Big data analytics and machine learning in hematology: Transformative insights, applications and challenges

Medicine, Journal Year: 2025, Volume and Issue: 104(10), P. e41766 - e41766

Published: March 7, 2025

The integration of big data analytics and machine learning (ML) into hematology has ushered in a new era precision medicine, offering transformative insights disease management. By leveraging vast diverse datasets, including genomic profiles, clinical laboratory results, imaging data, these technologies enhance diagnostic accuracy, enable robust prognostic modeling, support personalized therapeutic interventions. Advanced ML algorithms, such as neural networks ensemble learning, facilitate the discovery novel biomarkers refine risk stratification for hematological disorders, leukemias, lymphomas, coagulopathies. Despite advancements, significant challenges persist, particularly realms integration, algorithm validation, ethical concerns. heterogeneity datasets lack standardized frameworks complicate their application, while "black-box" nature models raises issues reliability trust. Moreover, safeguarding patient privacy an data-driven medicine remains paramount, necessitating development secure analytical practices. Addressing is critical to ensuring equitable effective implementation technologies. Collaborative efforts between hematologists, scientists, bioinformaticians are pivotal translating innovations real-world practice. Emphasis on developing explainable artificial intelligence models, integrating real-time analytics, adopting federated approaches will further utility adoption As continue evolve, potential revolutionize improve outcomes immense.

Language: Английский

---

Artificial intelligence-driven prediction of COVID-19-related hospitalization and death: a systematic review

Frontiers in Public Health, Journal Year: 2023, Volume and Issue: 11

Published: June 20, 2023

Aim To perform a systematic review on the use of Artificial Intelligence (AI) techniques for predicting COVID-19 hospitalization and mortality using primary secondary data sources. Study eligibility criteria Cohort, clinical trials, meta-analyses, observational studies investigating or artificial intelligence were eligible. Articles without full text available in English language excluded. Data sources recorded Ovid MEDLINE from 01/01/2019 to 22/08/2022 screened. extraction We extracted information sources, AI models, epidemiological aspects retrieved studies. Bias assessment A bias models was done PROBAST. Participants Patients tested positive COVID-19. Results included 39 related AI-based prediction death The articles published period 2019-2022, mostly used Random Forest as model with best performance. trained cohorts individuals sampled populations European non-European countries, cohort sample size <5,000. collection generally demographics, records, laboratory results, pharmacological treatments (i.e., high-dimensional datasets). In most studies, internally validated cross-validation, but majority lacked external validation calibration. Covariates not prioritized ensemble approaches however, still showed moderately good performances Area under Receiver operating characteristic Curve (AUC) values >0.7. According PROBAST, all had high risk and/or concern regarding applicability. Conclusions broad range have been predict mortality. reported performance applicability detected.

Language: Английский

---

Automatic ARDS surveillance with chest X-ray recognition using convolutional neural networks

Journal of Critical Care, Journal Year: 2024, Volume and Issue: 82, P. 154794 - 154794

Published: March 28, 2024

Language: Английский

---

A Transformer-Based Model Trained on Large Scale Claims Data for Prediction of Severe COVID-19 Disease Progression

IEEE Journal of Biomedical and Health Informatics, Journal Year: 2023, Volume and Issue: 27(9), P. 4548 - 4558

Published: June 24, 2023

In situations like the COVID-19 pandemic, healthcare systems are under enormous pressure as they can rapidly collapse burden of crisis. Machine learning (ML) based risk models could lift by identifying patients with a high severe disease progression. Electronic Health Records (EHRs) provide crucial sources information to develop these because rely on routinely collected data. However, EHR data is challenging for training ML it contains irregularly timestamped diagnosis, prescription, and procedure codes. For such data, transformer-based promising. We extended previously published Med-BERT model including age, sex, medications, quantitative clinical measures, state information. After pre-training approximately 988 million EHRs from 3.5 patients, we developed predict Acute Respiratory Manifestations (ARM) using medical history 80,211 patients. Compared Random Forests, XGBoost, RETAIN, our more accurately forecast developing ARM after infection. used Integrated Gradients Bayesian networks understand link between essential features model. Finally, evaluated adapting Austrian in-patient Our study highlights promise predictive precision medicine.

Language: Английский

---

Platelet aggregates detected using quantitative phase imaging associate with COVID-19 severity

Communications Medicine, Journal Year: 2023, Volume and Issue: 3(1)

Published: Nov. 7, 2023

The clinical spectrum of acute SARS-CoV-2 infection ranges from an asymptomatic to life-threatening disease. Considering the broad severity, reliable biomarkers are required for early risk stratification and prediction outcomes. Despite numerous efforts, no COVID-19-specific biomarker has been established guide further diagnostic or even therapeutic approaches, most likely due insufficient validation, methodical complexity, economic factors. COVID-19-associated coagulopathy is a hallmark disease mainly attributed dysregulated immunothrombosis. This process describes intricate interplay platelets, innate immune cells, coagulation cascade, vascular endothelium leading both micro- macrothrombotic complications. In this context, increased levels immunothrombotic components, including platelet platelet-leukocyte aggregates, have described linked COVID-19 severity.Here, we describe label-free quantitative phase imaging approach, allowing identification cell-aggregates their components at single-cell resolution within 30 min, which prospectively qualifies method as point-of-care (POC) testing.We find significant association between severity amount aggregates. Additionally, observe linkage aggregate composition, size distribution platelets in aggregates.This study presents POC-compatible rapid analysis blood cell aggregates patients with COVID-19.The human body produces series responses when it gets infected SARS-CoV-2, virus that causes COVID-19. One these involves clotting factor sticking cells form bloodstream. We aimed understand significance progression. A approach was used investigate number patient blood. observed severe associated higher numbers specific composition Our can potentially support prevent complications other medical disorders, where shown aggregate.

Language: Английский

---

A Machine Learning Model for Predicting Intensive Care Unit Admission in Inpatients with COVID-19 Using Clinical Data and Laboratory Biomarkers

Biomedicines, Journal Year: 2025, Volume and Issue: 13(5), P. 1025 - 1025

Published: April 24, 2025

Background: Artificial intelligence tools can help improve the clinical management of patients with severe COVID-19. The aim this study was to validate a machine learning model predict admission Intensive Care Unit (ICU) in individuals Methods: A total 201 hospitalized COVID-19 were included. Sociodemographic and data as well laboratory biomarker results obtained from medical records information system. Three models generated, trained, internally validated: logistic regression (LR), random forest (RF), extreme gradient boosting (XGBoost). evaluated for sensitivity (Sn), specificity (Sp), area under curve (AUC), precision (P), SHapley Additive exPlanation (SHAP) values, utility predictive using decision analysis (DCA). Results: included following variables: type 2 diabetes mellitus (T2DM), obesity, absolute neutrophil basophil counts, neutrophil-to-lymphocyte ratio (NLR), D-dimer levels on day hospital admission. LR showed an Sn 0.67, Sp 0.65, AUC 0.74, P 0.66. RF achieved 0.87, 0.83, 0.96, 0.85. XGBoost demonstrated 0.85, 0.95, 0.86. Conclusions: Among models, robust performance (Sn = 0.86) favorable net benefit analysis, confirming its suitability predicting ICU aiding decision-making.

Language: Английский

---

A comprehensive ML-based Respiratory Monitoring System for Physiological Monitoring & Resource Planning in the ICU

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Jan. 23, 2024

Abstract Acute hypoxemic respiratory failure (RF) occurs frequently in critically ill patients and is associated with substantial morbidity, mortality increased resource use. We used machine learning to create a comprehensive monitoring system assist intensive care unit (ICU) physicians managing acute RF. The encompasses early detection ongoing of RF, assessment readiness for tracheal extubation prediction the risk failure. In study patients, model predicted 80% RF events at precision 45%, 65% identified more than 10 hours before onset. System predictive performance was significantly higher standard clinical based on patient’s oxygenation index successfully validated an external cohort ICU patients. have demonstrated how estimated (EF) could facilitate prevention both, unnecessarily prolonged mechanical ventilation. Furthermore, we illustrated machine-learning-based risk, along necessity ventilation patient-by-patient basis, can planning ICU. Specifically, our ICU-level ventilator use within 8 16 into future, mean absolute error 0.4 ventilators per effective capacity.

Language: Английский

---

Artificial Intelligence in the Intensive Care Unit: Present and Future in the COVID-19 Era

Journal of Personalized Medicine, Journal Year: 2023, Volume and Issue: 13(6), P. 891 - 891

Published: May 25, 2023

The development of artificial intelligence (AI) allows for the construction technologies capable implementing functions that represent human mind, senses, and problem-solving skills, leading to automation, rapid data analysis, acceleration tasks. These solutions has been initially implemented in medical fields relying on image analysis; however, technological interdisciplinary collaboration introduction AI-based enhancements further specialties. During COVID-19 pandemic, novel established big analysis experienced a expansion. Yet, despite possibilities advancements with these AI technologies, there are number shortcomings need be resolved assert highest safest level performance, especially setting intensive care unit (ICU). Within ICU, numerous factors affect clinical decision making work management could managed by technologies. Early detection patient’s deterioration, identification unknown prognostic parameters, or even improvement organization few many areas where patients personnel can benefit from developed AI.

Language: Английский

---