Explainable Machine Learning Model for Predicting Persistent Sepsis-Associated Acute Kidney Injury: Development and Validation Study (Preprint) DOI
Wei Jiang,

Yueyue Zhang,

Jiayi Weng

et al.

Published: June 5, 2024

BACKGROUND Persistent sepsis-associated acute kidney injury (SA-AKI) shows poor clinical outcomes and remains a therapeutic challenge for clinicians. Early identification prediction of persistent SA-AKI are crucial. OBJECTIVE The aim this study was to develop validate an interpretable machine learning (ML) model that predicts compare its diagnostic performance with C-C motif chemokine ligand 14 (CCL14) in prospective cohort. METHODS used 4 retrospective cohorts 1 cohort derivation validation. the MIMIC-IV database, which randomly split into 2 parts (80% construction 20% internal validation). External validation conducted using subsets MIMIC-III dataset e-ICU dataset, from intensive care unit (ICU) Northern Jiangsu People’s Hospital. Prospective data same ICU were comparison urinary CCL14 biomarker measurements. Acute (AKI) defined based on serum creatinine urine output, Kidney Disease: Improving Global Outcomes (KDIGO) criteria. Routine within first 24 hours admission collected, 8 ML algorithms construct model. Multiple evaluation metrics, including area under receiver operating characteristic curve (AUC), predictive performance. Feature importance ranked Shapley Additive Explanations (SHAP), final explained accordingly. In addition, developed web-based application Streamlit framework facilitate application. RESULTS A total 46,097 patients sepsis multiple enrolled analysis. Among 17,928 cohort, 8081 (45.1%) showed progression SA-AKI. models, gradient boosting (GBM) demonstrated superior discriminative ability. Following feature ranking, GBM comprising 12 features (AKI stage, ΔCreatinine, furosemide dose, BMI, Sequential Organ Failure Assessment score, replacement therapy, mechanical ventilation, lactate, blood urea nitrogen, prothrombin time, age) established. accurately predicted occurrence both (AUC=0.870) external (MIMIC-III subset: AUC=0.891; dataset: AUC=0.932; Hospital cohort: AUC=0.983). outperformed predicting (GBM AUC=0.852 vs AUC=0.821). has been transformed online tool settings. CONCLUSIONS shown successfully predict SA-AKI, demonstrating good ability cohorts. Furthermore, outperform

Language: Английский

Molybdenum Disulfide-Supported Cuprous Oxide Nanocomposite for Near-Infrared-I Light-Responsive Synergistic Antibacterial Therapy DOI
Jiao Li, Jie Li, Yuli Chen

et al.

ACS Nano, Journal Year: 2024, Volume and Issue: 18(25), P. 16184 - 16198

Published: June 12, 2024

Drug-resistant bacterial infections pose a serious threat to human health; thus, there is an increasingly growing demand for nonantibiotic strategies overcome drug resistance in infections. Mild photothermal therapy (PTT), as attractive antibacterial strategy, shows great potential application due its good biocompatibility and ability circumvent resistance. However, efficiency limited by the heat of bacteria. Herein, Cu2O@MoS2, nanocomposite, was constructed situ growth Cu2O nanoparticles (NPs) on surface MoS2 nanosheets, which provided controllable therapeutic effect intrinsic catalytic properties NPs, achieving synergistic eradicate multidrug-resistant Transcriptome sequencing (RNA-seq) results revealed that process related disrupting membrane transport system, phosphorelay signal transduction oxidative stress response well system. Animal experiments indicated Cu2O@MoS2 could effectively treat wounds infected with methicillin-resistant Staphylococcus aureus. In addition, satisfactory made promising agent. Overall, our highlight nanocomposite solution combating resistant bacteria without inducing evolution antimicrobial

Language: Английский

Citations

20
The Role of Artificial Intelligence and Machine Learning in Predicting and Combating Antimicrobial Resistance DOI Creative Commons
Hazrat Bilal, Muhammad Nadeem Khan, Sabir Khan

et al.

Computational and Structural Biotechnology Journal, Journal Year: 2025, Volume and Issue: 27, P. 423 - 439

Published: Jan. 1, 2025

Antimicrobial resistance (AMR) is a major threat to global public health. The current review synthesizes address the possible role of Artificial Intelligence and Machine Learning (AI/ML) in mitigating AMR. Supervised learning, unsupervised deep reinforcement natural language processing are some main tools used this domain. AI/ML models can use various data sources, such as clinical information, genomic sequences, microbiome insights, epidemiological for predicting AMR outbreaks. Although relatively new fields, numerous case studies offer substantial evidence their successful application outbreaks with greater accuracy. These provide insights into discovery novel antimicrobials, repurposing existing drugs, combination therapy through analysis molecular structures. In addition, AI-based decision support systems real-time guide healthcare professionals improve prescribing antibiotics. also outlines how AI surveillance, analyze trends, enable early outbreak identification. Challenges, ethical considerations, privacy, model biases exist, however, continuous development methodologies enables play significant combating

Language: Английский

Citations

5
The Role of Artificial Intelligence and Machine Learning Models in Antimicrobial Stewardship in Public Health: A Narrative Review DOI Creative Commons
Flavia Pennisi,

A.C. Pinto,

Giovanni Emanuele Ricciardi

et al.

Antibiotics, Journal Year: 2025, Volume and Issue: 14(2), P. 134 - 134

Published: Jan. 30, 2025

Antimicrobial resistance (AMR) poses a critical global health threat, necessitating innovative approaches in antimicrobial stewardship (AMS). Artificial intelligence (AI) and machine learning (ML) have emerged as transformative tools this domain, enabling data-driven interventions to optimize antibiotic use combat resistance. This comprehensive review explores the multifaceted role of AI ML models enhancing efforts across healthcare systems. AI-powered predictive analytics can identify patterns resistance, forecast outbreaks, guide personalized therapies by leveraging large-scale clinical epidemiological data. algorithms facilitate rapid pathogen identification, profiling, real-time monitoring, precise decision making. These technologies also support development advanced diagnostic tools, reducing reliance on broad-spectrum antibiotics fostering timely, targeted treatments. In public health, AI-driven surveillance systems improve detection AMR trends enhance monitoring capabilities. By integrating diverse data sources—such electronic records, laboratory results, environmental data—ML provide actionable insights policymakers, providers, officials. Additionally, applications programs (ASPs) promote adherence prescribing guidelines, evaluate intervention outcomes, resource allocation. Despite these advancements, challenges such quality, algorithm transparency, ethical considerations must be addressed maximize potential field. Future research should focus developing interpretable interdisciplinary collaborations ensure equitable sustainable integration into initiatives.

Language: Английский

Citations

3
Artificial intelligence in antimicrobial stewardship: a systematic review and meta-analysis of predictive performance and diagnostic accuracy DOI
Flavia Pennisi, Armando Teixeira‐Pinto, Giovanni Emanuele Ricciardi

et al.

European Journal of Clinical Microbiology & Infectious Diseases, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 6, 2025

Language: Английский

Citations

1
Global insights into MRSA bacteremia: a bibliometric analysis and future outlook DOI Creative Commons
Jiayi Lin, Jia-Kai Lai, Jianyi Chen

et al.

Frontiers in Microbiology, Journal Year: 2025, Volume and Issue: 15

Published: Jan. 22, 2025

Background Methicillin-resistant Staphylococcus aureus (MRSA) bloodstream infections (BSIs) pose a significant challenge to global public health, characterized by high morbidity and mortality rates, particularly in immunocompromised patients. Despite extensive research, the rapid development of MRSA antibiotic resistance has outpaced current treatment methods, increasing difficulty treatment. Therefore, reviewing research on BSIs is crucial. Methods This study conducted bibliometric analysis, retrieving analyzing 1,621 publications related from 2006 2024. The literature was sourced Web Science Core Collection (WoSCC), data visualization trend analysis were performed using VOSviewer, CiteSpace, Bibliometrix software packages. Results showed that primarily concentrated United States, China, Japan. States leads output influence, with contributions institutions such as University California system Texas system. journal most Antimicrobial Agents Chemotherapy, while cited publication Vincent JL’s article “Sepsis European Intensive Care Units: SOAP Study” published Critical Medicine 2006. Cosgrove SE’s “Comparison Mortality Associated Methicillin-Resistant Methicillin-Susceptible Bacteremia: A Meta-analysis” had co-citations. Key trends include MRSA’s mechanisms, application new diagnostic technologies, impact COVID-19 studies. Additionally, artificial intelligence (AI) machine learning are increasingly applied diagnosis treatment, phage therapy vaccine have become future hotspots. Conclusion remain major health challenge, especially severity resistance. Although progress been made treatments further validation required. Future will rely integrating genomics, AI, drive personalized Strengthening cooperation, resource-limited countries, be key effectively addressing BSIs.

Language: Английский

Citations

1
Antibiotics and Artificial Intelligence: Clinical Considerations on a Rapidly Evolving Landscape DOI Creative Commons
Daniele Roberto Giacobbe, Sabrina Guastavino, Cristina Marelli

et al.

Infectious Diseases and Therapy, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 15, 2025

The growing interest in leveraging artificial intelligence (AI) tools for healthcare decision-making extends to improving antibiotic prescribing. Large language models (LLMs), a type of AI trained on extensive datasets from diverse sources, can process and generate contextually relevant text. While their potential enhance patient outcomes is significant, implementing LLM-based support prescribing complex. Here, we specifically expand the discussion this crucial topic by introducing three interconnected perspectives: (1) distinctive commonalities, but also conceptual differences, between use LLMs as assistants scientific writing supporting real-world practice; (2) possibility nuances expertise paradox; (3) peculiarities risk error when considering complex tasks such

Language: Английский

Citations

0
MOF nanozyme mediated bacterial metabolic regulation to intervene MRSA antibiotic tolerance for enhanced antimicrobial efficacy DOI
Mingkai Wang, Ruiyang Li, Shihao Sheng

et al.

Nano Today, Journal Year: 2025, Volume and Issue: 63, P. 102753 - 102753

Published: April 10, 2025

Language: Английский

Citations

0
Knowledge graph driven medicine recommendation system using graph neural networks on longitudinal medical records DOI Creative Commons

Rajat Mishra,

S. Shridevi

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

Published: Oct. 26, 2024

Abstract Medicine recommendation systems are designed to aid healthcare professionals by analysing a patient’s admission data recommend safe and effective medications. These categorised into two types: instance-based longitudinal-based. Instance-based models only consider the current admission, while longitudinal medical history. Electronic Health Records used incorporate history models. This project proposes novel K nowledge G raph- D riven Recommendation System using Graph Neural Net works, KGDNet , that utilises EHR along with ontologies Drug-Drug Interaction knowledge construct admission-wise clinical medicine Knowledge Graphs for every patient. Recurrent Networks employed model historical data, learn embeddings from Graphs. A Transformer-based Attention mechanism is then generate medication recommendations patient, considering their state, history, joint records. The evaluated on MIMIC-IV outperforms existing methods in terms of precision, recall, F1 score, Jaccard control. An ablation study our various inputs components provide evidence importance each component providing best performance. Case also performed demonstrate real-world effectiveness KGDNet.

Language: Английский

Citations

3
Learning-based early detection of post-hepatectomy liver failure using temporal perioperative data: a nationwide multicenter retrospective study in China DOI Creative Commons
Kai Wang, Qian Yang, Kang Li

et al.

EClinicalMedicine, Journal Year: 2025, Volume and Issue: 83, P. 103220 - 103220

Published: May 1, 2025

Language: Английский

Citations

0
Generative AI models in time varying biomedical data: a systematic review (Preprint) DOI Creative Commons
Rosemary He, Varuni Sarwal, Xinru Qiu

et al.

Published: April 30, 2024

BACKGROUND Trajectory modeling is a longstanding challenge in the application of computational methods to healthcare. However, traditional statistical and machine learning do not achieve satisfactory results as they often fail capture complex underlying distributions multi-modal health data, long-term dependencies throughout patients’ medical histories. Recent advances generative AI have provided powerful tools represent patterns with minimal assumptions. These had major impact fields such finance environmental sciences, recently researchers turned these for disease modeling. OBJECTIVE While proven powerful, their clinical practice remains limited due black-box like nature. The proliferation algorithms poses significant non-developers track incorporate into research application. In this work, we survey peer-reviewed, model papers specific applications time series data. METHODS Our search includes single- models that operate over structured unstructured imaging or multi-omics We introduce current methods, review each data modality discuss strengths weaknesses compared methods. RESULTS follow PRISMA guideline 155 articles on healthcare across modalities. Furthermore, offer systematic framework clinicians easily identify suitable task at hand. CONCLUSIONS critique existing aim bridging gap between also shortcomings approaches, highlight recent promising directions

Language: Английский

Citations

1