Machine learning for clinical decision support in infectious diseases: a narrative review of current applications

Modeling the determinants of attrition in a two-stage epilepsy prevalence survey in Nairobi using machine learning

Background

Attrition is a challenge in parameter estimation in both longitudinal and multi-stage cross-sectional studies. Here, we examine utility of machine learning to predict attrition and identify associated factors in a two-stage population-based epilepsy prevalence study in Nairobi.

Methods

All individuals in the Nairobi Urban Health and Demographic Surveillance System (NUHDSS) (Korogocho and Viwandani) were screened for epilepsy in two stages. Attrition was defined as probable epilepsy cases identified at stage-I but who did not attend stage-II (neurologist assessment). Categorical variables were one-hot encoded, class imbalance was addressed using synthetic minority over-sampling technique (SMOTE) and numeric variables were scaled and centered. The dataset was split into training and testing sets (7:3 ratio), and seven machine learning models, including the ensemble Super Learner, were trained. Hyperparameters were tuned using 10-fold cross-validation, and model performance evaluated using metrics like Area under the curve (AUC), accuracy, Brier score and F1 score over 500 bootstrap samples of the test data.

Results

Random forest (AUC = 0.98, accuracy = 0.95, Brier score = 0.06, and F1 = 0.94), extreme gradient boost (XGB) (AUC = 0.96, accuracy = 0.91, Brier score = 0.08, F1 = 0.90) and support vector machine (SVM) (AUC = 0.93, accuracy = 0.93, Brier score = 0.07, F1 = 0.92) were the best performing models (base learners). Ensemble Super Learner had similarly high performance. Important predictors of attrition included proximity to industrial areas, male gender, employment, education, smaller households, and a history of complex partial seizures.

Conclusion

These findings can aid researchers plan targeted mobilization for scheduled clinical appointments to improve follow-up rates. These findings will inform development of a web-based algorithm to predict attrition risk and aid in targeted follow-up efforts in similar studies.

Deep learning models for early and accurate diagnosis of ventilator-associated pneumonia in mechanically ventilated neonates

BACKGROUND

Early and accurate confirmation of critically ill neonates with a suspected diagnosis of ventilator-associated pneumonia (VAP) can optimize the therapeutic strategy and avoid unnecessary use of empirical antibiotics. We aimed to examine whether deep learning (DL) methods can assist the diagnosis of VAP of intubated neonates in the neonatal intensive care unit (NICU).

METHODS

A total of 670 neonates with mechanical ventilation were prospectively observed in a tertiary-level NICU in Taiwan between October 2017 and March 2022, during which image data were collected. All neonates with clinically suspected VAP were enrolled, and various DL methods were used to test the prediction ability of VAP diagnosis. The accuracy, precision, sensitivity, specificity, F1-score, and area under curves (AUCs) of several DL methods were compared.

RESULTS

A total of 900 chest X-ray images derived from 670 neonates with VAP and/or bronchopulmonary dysplasia (BPD) were enrolled, including 399 images from patients with definite diagnosis of VAP based on the strict criteria and 501 images from neonates without VAP. Compared with conventional DNN models such as ResNet, VGG, DenseNet, the RegNetX80 achieved the best specificity of 0.8378, which facilitates a low false positive rate. For accurate diagnosis of neonatal VAP, a combinatorial model of ResNet50 and RegNetX80, created through ensemble learning, further enhanced the AUC to 0.8023 for neonates with VAP on mechanical ventilation. In addition, the consistent XAI results in the left-lower region of chest X-ray image provided informative feedback and increased confidence to AI-assisted doctors.

CONCLUSIONS

Deep learning methods are applicable with good predictive accuracy using chest X-ray images to help diagnosis of VAP in the NICU, which can help clinicians make decisions regarding the choices of empiric antibiotics for critically ill neonates. Future prospective trials are warranted to document its clinical usefulness and benefits on reducing medical resources.

Application of machine learning models to explore prognosis and cause of death in advanced intrahepatic cholangiocarcinoma patients undergoing chemotherapy

BACKGROUND

This study was aimed at examining the causes of death (CODs) in patients with advanced intrahepatic cholangiocarcinoma (ICC) undergoing chemotherapy (CT). In addition, machine learning models were incorporated to predict the treatment outcomes of patients with advanced ICC and identify the factors most closely related to prognosis.

METHODS

A total of 5564 patients (CT group, 3632; non-CT group, 1932) were included in the Surveillance Epidemiology and End Results registries between 2000 and 2020. The CODs were compared between the two groups before and after the inverse probability of treatment weighting (IPTW). Furthermore, seven machine learning models were utilized as predictive tools to select variable features, aiming to assess the therapeutic effectiveness in patients with advanced ICC.

RESULTS

After IPTW, the CT group exhibited a lower cumulative incidence of cholangiocarcinoma-related deaths (30%, 49%, and 73% vs. 59%, 66%, and 73%; P < 0.001), secondary malignant neoplasms (8.5%, 13%, and 20% vs. 19%, 22%, and 24%; P < 0.001), and other CODs (1.8%, 2.9%, and 4.4% vs. 4.1%, 4.6%, and 5.4%; P < 0.001) at 0.5-, 1-, and 3- years than the non-CT group, whereas the cumulative incidence of cardiovascular diseases (P = 0.4) was comparable between the two groups. Of the seven machine learning models, the random forest model showed the highest predictive effectiveness. This model verified that variables such as CT, radiotherapy, tumor dimensions, sex, and distant metastasis were strongly correlated with the prognosis of advanced ICC.

CONCLUSIONS

CT has improved the therapeutic efficacy of advanced ICC without significantly increasing other CODs. Furthermore, the analysis of various features using machine learning models has confirmed that the random forest model demonstrates the highest predictive performance.

Integrating socioeconomic deprivation indices and electronic health record data to predict antimicrobial resistance

We developed machine learning models to predict the presence of AMR organisms in blood cultures obtained at the first patient encounter, offering a new and inspiring direction for antimicrobial resistance management. Three supervised machine learning classifiers were used: penalized logistic regression, random forest, and XGBoost, which were used to classify five AMR organisms: ESBL, CRE, AmpC, MRSA, and VRE. The random forest and XGBoost models performed best, with AUC-ROC values of 0.70 and 92.9% negative predictive value, respectively. The multi-class random forest model's AUC-ROC values ranged from 0.80-0.95. Our models highlight how the combination of ADI and SVI increased the predictive power. This approach could reduce costs and mitigate the global public health threat posed by antibiotic-resistant infections. Machine learning techniques can predict antimicrobial-resistant infections in suspected cultures using patient data from EHRs, enabling clinicians to make targeted prescribing decisions and mitigate resistance development.

Machine learning-based model for predicting all-cause mortality in severe pneumonia

BACKGROUND

Severe pneumonia has a poor prognosis and high mortality. Current severity scores such as Acute Physiology and Chronic Health Evaluation (APACHE-II) and Sequential Organ Failure Assessment (SOFA), have limited ability to help clinicians in classification and management decisions. The goal of this study was to analyse the clinical characteristics of severe pneumonia and develop a machine learning-based mortality-prediction model for patients with severe pneumonia.

METHODS

Consecutive patients with severe pneumonia between 2013 and 2022 admitted to Beijing Chaoyang Hospital affiliated with Capital Medical University were included. In-hospital all-cause mortality was the outcome of this study. We performed a retrospective analysis of the cohort, stratifying patients into survival and non-survival groups, using mainstream machine learning algorithms (light gradient boosting machine, support vector classifier and random forest). We aimed to construct a mortality-prediction model for patients with severe pneumonia based on their accessible clinical and laboratory data. The discriminative ability was evaluated using the area under the receiver operating characteristic curve (AUC). The calibration curve was used to assess the fit goodness of the model, and decision curve analysis was performed to quantify clinical utility. By means of logistic regression, independent risk factors for death in severe pneumonia were figured out to provide an important basis for clinical decision-making.

RESULTS

A total of 875 patients were included in the development and validation cohorts, with the in-hospital mortality rate of 14.6%. The AUC of the model in the internal validation set was 0.8779 (95% CI, 0.738 to 0.974), showing a competitive discrimination ability that outperformed those of traditional clinical scoring systems, that is, APACHE-II, SOFA, CURB-65 (confusion, urea, respiratory rate, blood pressure, age ≥65 years), Pneumonia Severity Index. The calibration curve showed that the in-hospital mortality in severe pneumonia predicted by the model fit reasonably with the actual hospital mortality. In addition, the decision curve showed that the net clinical benefit was positive in both training and validation sets of hospitalised patients with severe pneumonia. Based on ensemble machine learning algorithms and logistic regression technique, the level of ferritin, lactic acid, blood urea nitrogen, creatine kinase, eosinophil and the requirement of vasopressors were identified as top independent predictors of in-hospital mortality with severe pneumonia.

CONCLUSION

A robust clinical model for predicting the risk of in-hospital mortality after severe pneumonia was successfully developed using machine learning techniques. The performance of this model demonstrates the effectiveness of these techniques in creating accurate predictive models, and the use of this model has the potential to greatly assist patients and clinical doctors in making well-informed decisions regarding patient care.

Using machine learning to predict depression among middle-aged and elderly population in China and conducting empirical analysis

OBJECTIVE

To develop a predictive model for evaluating depression among middle-aged and elderly individuals in China.

METHODS

Participants aged ≥ 45 from the 2020 China Health and Retirement Survey (CHARLS) cross-sectional study were enrolled. Depressive mood was defined as a score of 10 or higher on the CESD-10 scale, which has a maximum score of 30. A predictive model was developed using five selected machine learning algorithms. The model was trained and validated on the 2020 database cohort and externally validated through a questionnaire survey of middle-aged and elderly individuals in Shaanxi Province, China, following the same criteria. SHapley Additive Interpretation (SHAP) was employed to assess the importance of predictive factors.

RESULTS

The stacked ensemble model demonstrated an AUC of 0.8021 in the test set of the training cohort for predicting depressive symptoms; the corresponding AUC in the external validation cohort was 0.7448, outperforming all base models.

CONCLUSION

The stacked ensemble approach serves as an effective tool for identifying depression in a large population of middle-aged and elderly individuals in China. For depression prediction, factors such as life satisfaction, self-reported health, pain, sleep duration, and cognitive function are identified as highly significant predictive factors.

Comprehensive integration of diagnostic biomarker analysis and immune cell infiltration features in sepsis via machine learning and bioinformatics techniques

Introduction

Sepsis, a critical medical condition resulting from an irregular immune response to infection, leads to life-threatening organ dysfunction. Despite medical advancements, the critical need for research into dependable diagnostic markers and precise therapeutic targets.

Methods

We screened out five gene expression datasets (GSE69063, GSE236713, GSE28750, GSE65682 and GSE137340) from the Gene Expression Omnibus. First, we merged the first two datasets. We then identified differentially expressed genes (DEGs), which were subjected to KEGG and GO enrichment analyses. Following this, we integrated the DEGs with the genes from key modules as determined by Weighted Gene Co-expression Network Analysis (WGCNA), identifying 262 overlapping genes. 12 core genes were subsequently selected using three machine-learning algorithms: random forest (RF), Least Absolute Shrinkage and Selection Operator (LASSO), and Support Vector Machine-Recursive Feature Elimination (SVW-RFE). The utilization of the receiver operating characteristic curve in conjunction with the nomogram model served to authenticate the discriminatory strength and efficacy of the key genes. CIBERSORT was utilized to evaluate the inflammatory and immunological condition of sepsis. Astragalus, Salvia, and Safflower are the primary elements of Xuebijing, commonly used in the clinical treatment of sepsis. Using the Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP), we identified the chemical constituents of these three herbs and their target genes.

Results

We found that CD40LG is not only one of the 12 core genes we identified, but also a common target of the active components quercetin, luteolin, and apigenin in these herbs. We extracted the common chemical structure of these active ingredients -flavonoids. Through docking analysis, we further validated the interaction between flavonoids and CD40LG. Lastly, blood samples were collected from healthy individuals and sepsis patients, with and without the administration of Xuebijing, for the extraction of peripheral blood mononuclear cells (PBMCs). By qPCR and WB analysis. We observed significant differences in the expression of CD40LG across the three groups. In this study, we pinpointed candidate hub genes for sepsis and constructed a nomogram for its diagnosis.

Discussion

This research not only provides potential diagnostic evidence for peripheral blood diagnosis of sepsis but also offers insights into the pathogenesis and disease progression of sepsis.

A Critical Review of the Prospect of Integrating Artificial Intelligence in Infectious Disease Diagnosis and Prognosis

This paper explores the transformative potential of integrating artificial intelligence (AI) in the diagnosis and prognosis of infectious diseases. By analyzing diverse datasets, including clinical symptoms, laboratory results, and imaging data, AI algorithms can significantly enhance early detection and personalized treatment strategies. This paper reviews how AI-driven models improve diagnostic accuracy, predict patient outcomes, and contribute to effective disease management. It also addresses the challenges and ethical considerations associated with AI, including data privacy, algorithmic bias, and equitable access to healthcare. Highlighting case studies and recent advancements, the paper underscores AI's role in revolutionizing infectious disease management and its implications for future healthcare delivery.

Interpretable machine learning model for early morbidity risk prediction in patients with sepsis-induced coagulopathy: a multi-center study

Background

Sepsis-induced coagulopathy (SIC) is a complex condition characterized by systemic inflammation and coagulopathy. This study aimed to develop and validate a machine learning (ML) model to predict SIC risk in patients with sepsis.

Methods

Patients with sepsis admitted to the intensive care unit (ICU) between March 1, 2021, and March 1, 2024, at Hebei General Hospital and Handan Central Hospital (East District) were retrospectively included. Patients were categorized into SIC and non-SIC groups. Data were split into training (70%) and testing (30%) sets. Additionally, for temporal validation, patients with sepsis admitted between March 1, 2024, and October 31, 2024, at Hebei General Hospital were included. Feature selection was performed using least absolute shrinkage and selection operator (LASSO) regression and multivariate logistic regression. Nine ML algorithms were tested, and model performance was assessed using receiver operating characteristic curve (ROC) analysis, including area under the curve (AUC), calibration curves, and decision curve analysis (DCA). The SHaply Additive Explanations (SHAP) algorithm was used to interpret the best-performing model and visualize key predictors.

Results

Among 847 patients with sepsis, 480 (56.7%) developed SIC. The random forest (RF) model with eight variables performed best, achieving AUCs of 0.782 [95% confidence interval (CI): 0.745, 0.818] in the training set, 0.750 (95% CI: 0.690, 0.809) in the testing set, and 0.784 (95% CI: 0.711, 0.857) in the validation set. Key predictors included activated partial thromboplastin time, lactate, oxygenation index, and total protein.

Conclusions

This ML model reliably predicts SIC risk. SHAP enhances interpretability, supporting early, individualized interventions to improve outcomes in patients with sepsis.

Utilising routinely collected clinical data through time series deep learning to improve identification of bacterial bloodstream infections: a retrospective cohort study

BACKGROUND

Blood cultures are the gold standard for diagnosing bacterial bloodstream infections, but test results are only available 24-48 h after sampling. We aimed to develop and evaluate models using health-care data to predict bloodstream infections in patients admitted to hospital.

METHODS

In this retrospective cohort study, we used routinely collected blood biomarkers and demographic data from patients who underwent blood sample collection for testing via culture between March 3, 2014, and Dec 1, 2021, at Imperial College Healthcare NHS Trust (London, UK) as model features. Data up to 14 days before blood sample collection were provided to long short-term memory (LSTM) or static logistic regression models. The primary outcome was prediction of blood culture results, defined as a pathogenic bloodstream infection (ie, isolation of pathogenic bacteria of interest) or no bloodstream infection (ie, no growth or contamination). Data collected up to Feb 28, 2021 (n=15 212) comprised the training set and were evaluated against a temporal hold-out test set comprising patients who were sampled after March 1, 2021 (n=5638).

FINDINGS

Among 20 850 patients with available data, pathogenic bacteria were observed in the cultured blood samples of 3866 (18·5%) patients. 2920 (62·2%) of 4897 patients who had their blood samples taken more than 48 h after admission to hospital had pathogenic bloodstream infections, and so were defined as having hospital-acquired bloodstream infections. Including data from the 7 days before admission (7-day window approach) and using five-fold cross validation in the training set gave an area under receiver operator curve (AUROC) of 0·75 (IQR 0·68-0·82) and an area under the precision recall curve (AUPRC) of 0·58 (0·46-0·77) for static models and an AUROC of 0·92 (0·91-0·93) and AUPRC of 0·75 (0·72-0·76) for the LSTM model. In the hold-out test set performances were: AUROC of 0·74 (95% CI 0·70-0·78) and AUPRC of 0·48 (0·43-0·53) for static models and AUROC of 0·97 (0·96-0·97) and AUPRC of 0·65 (0·60-0·70) for LSTM. Removal of time series information resulted in lower model performance, particularly for hospital-acquired bloodstream infections. Dynamics of C-reactive protein concentration, eosinophil count, and platelet count were important features for prediction of blood culture results.

INTERPRETATION

Deep learning models accounting for longitudinal changes could support individualised clinical decision making for patients at risk of bloodstream infections. Appropriate implementation into existing diagnostic pathways could enhance diagnostic stewardship and reduce unnecessary antimicrobial prescribing.

FUNDING

UK Department of Health and Social Care, the National Institute for Health and Care Research, and the Wellcome Trust.

A Machine Learning Model for Diagnosing Opportunistic Infections in HIV Patients: Broad Applicability Across Infection Types

Opportunistic infections (OIs) are the leading cause of hospitalisation and mortality among Human Immunodeficiency Virus-infected (HIV-infected) patients. The diverse pathogen types and intricate clinical manifestations associated present a formidable challenge to the timely diagnosis of these infections. This study aims to use machine learning techniques to develop a diagnostic model that quickly identifies whether HIV-infected patients have any type of OIs, without being limited to specific infections, thus adapting to various clinical scenarios. This study is a retrospective cohort study that collected clinical data from HIV-infected patients at four healthcare organisations in China. A total of twelve machine learning classification algorithms were employed for the purposes of model training and evaluation. Additionally, feature reduction was conducted through the implementation of an importance ranking, with the objective of eliminating any redundant features. In conclusion, both the five features based on Shapley additive explanations (procalcitonin, haemoglobin, lymphocyte, creatinine, platelet) and the five features based on Permutation Importance explanations (procalcitonin, lymphocyte, haemoglobin, creatinine, indirect bilirubin) achieved the highest F1 score when evaluated using the adaptive boosting classifier model. The scores on the test set were 0.9016 and 0.9063, respectively, which significantly outperformed the best 32-feature model, gradient boosting classifier, which had a test set F1 score of 0.8991.

Identification of UBE2N as a biomarker of Alzheimer's disease by combining WGCNA with machine learning algorithms

Alzheimer's disease (AD) is the most common cause of dementia, emphasizing the critical need for the development of biomarkers that facilitate accurate and objective assessment of disease progression for early detection and intervention to delay its onset. In our study, three AD datasets from the Gene Expression Omnibus (GEO) database were integrated for differential expression analysis, followed by a weighted gene co-expression network analysis (WGCNA), and potential AD biomarkers were screened. Our study identified UBE2N as a promising biomarker for AD. Functional enrichment analysis revealed that UBE2N is associated with synaptic vesicle cycling and T cell/B cell receptor signaling pathways. Notably, UBE2N expression levels were found to be significantly reduced in the cortex and hippocampus of the TauP301S mice. Furthermore, analysis of single-cell data from AD patients demonstrated the association of UBE2N and T cell function. These findings underscore the potential of UBE2N as a valuable biomarker for AD, offering important insights for diagnosis and targeted therapeutic strategies.

Clinical validation and optimization of machine learning models for early prediction of sepsis

Introduction

Sepsis is a global health threat that has a high incidence and mortality rate. Early prediction of sepsis onset can drive effective interventions and improve patients' outcome.

Methods

Data were collected retrospectively from a cohort of 2,329 adult patients with positive bacteria cultures from a tertiary hospital in China between October 1, 2019 and September 30, 2020. Thirty six clinical features were selected as inputs for the models. We trained models in predicting sepsis by machine learning (ML) methods, including logistic regression, decision tree, random forest (RF), multi-layer perceptron, and light gradient boosting. We evaluated the performance of the five ML models and the evaluation metrics were: area under the ROC curve (AUC), accuracy, F1-score, sensitivity and specificity. The data of another cohort of 2,286 patients between October 1, 2020 and April 1, 2022 were used to validate the performance of the model performing best in the in the internal validation set. Shapley additive explanations (SHAP) method was applied to evaluate feature importance and explain the predictions of this model.

Results

Of the five machine learning models developed, the RF model demonstrated the best performance in terms of AUC (0.818), F1 value (0.38), and sensitivity (0.746). The RF model also has a comparable AUC (0.771) in the external validation set. The SHAP method identified procalcitonin, albumin, prothrombin time, and sex as the important variables contributing to the prediction of sepsis.

Discussion

The RF model we developed showed the greatest potential for early prediction of sepsis in admitted patients, which could aid clinicians in their decision-making process. Our findings also suggested that male patients with bacterial infections and high procalcitonin levels, lower albumin levels, or prolonged prothrombin times were more likely to develop sepsis.

The multiple uses of artificial intelligence in exercise programs: a narrative review

Background

Artificial intelligence is based on algorithms that enable machines to perform tasks and activities that generally require human intelligence, and its use offers innovative solutions in various fields. Machine learning, a subset of artificial intelligence, concentrates on empowering computers to learn and enhance from data autonomously; this narrative review seeks to elucidate the utilization of artificial intelligence in fostering physical activity, training, exercise, and health outcomes, addressing a significant gap in the comprehension of practical applications.

Methods

Only Randomized Controlled Trials (RCTs) published in English were included. Inclusion criteria: all RCTs that use artificial intelligence to program, supervise, manage, or assist physical activity, training, exercise, or health programs. Only studies published from January 1, 2014, were considered. Exclusion criteria: all the studies that used robot-assisted, robot-supported, or robotic training were excluded.

Results

A total of 1772 studies were identified. After the first stage, where the duplicates were removed, 1,004 articles were screened by title and abstract. A total of 24 studies were identified, and finally, after a full-text review, 15 studies were identified as meeting all eligibility criteria for inclusion. The findings suggest that artificial intelligence holds promise in promoting physical activity across diverse populations, including children, adolescents, adults, older adult, and individuals with disabilities.

Conclusion

Our research found that artificial intelligence, machine learning and deep learning techniques were used: (a) as part of applications to generate automatic messages and be able to communicate with users; (b) as a predictive approach and for gesture and posture recognition; (c) as a control system; (d) as data collector; and (e) as a guided trainer.

User-Oriented Requirements for Artificial Intelligence-Based Clinical Decision Support Systems in Sepsis: Protocol for a Multimethod Research Project

BACKGROUND

Artificial intelligence (AI)-based clinical decision support systems (CDSS) have been developed for several diseases. However, despite the potential to improve the quality of care and thereby positively impact patient-relevant outcomes, the majority of AI-based CDSS have not been adopted in standard care. Possible reasons for this include barriers in the implementation and a nonuser-oriented development approach, resulting in reduced user acceptance.

OBJECTIVE

This research project has 2 objectives. First, problems and corresponding solutions that hinder or support the development and implementation of AI-based CDSS are identified. Second, the research project aims to increase user acceptance by creating a user-oriented requirement profile, using the example of sepsis.

METHODS

The research project is based on a multimethod approach combining (1) a scoping review, (2) focus groups with physicians and professional caregivers, and (3) semistructured interviews with relevant stakeholders. The research modules mentioned provide the basis for the development of a (4) survey, including a discrete choice experiment (DCE) with physicians. A minimum of 6667 physicians with expertise in the clinical picture of sepsis are contacted for this purpose. The survey is followed by the development of a requirement profile for AI-based CDSS and the derivation of policy recommendations for action, which are evaluated in a (5) expert roundtable discussion.

RESULTS

The multimethod research project started in November 2022. It provides an overview of the barriers and corresponding solutions related to the development and implementation of AI-based CDSS. Using sepsis as an example, a user-oriented requirement profile for AI-based CDSS is developed. The scoping review has been concluded and the qualitative modules have been subjected to analysis. The start of the survey, including the DCE, was at the end of July 2024.

CONCLUSIONS

The results of the research project represent the first attempt to create a comprehensive user-oriented requirement profile for the development of sepsis-specific AI-based CDSS. In addition, general recommendations are derived, in order to reduce barriers in the development and implementation of AI-based CDSS. The findings of this research project have the potential to facilitate the integration of AI-based CDSS into standard care in the long term.

INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID)

DERR1-10.2196/62704.

Evolution of Artificial Intelligence in Medical Education From 2000 to 2024: Bibliometric Analysis

BACKGROUND

Incorporating artificial intelligence (AI) into medical education has gained significant attention for its potential to enhance teaching and learning outcomes. However, it lacks a comprehensive study depicting the academic performance and status of AI in the medical education domain.

OBJECTIVE

This study aims to analyze the social patterns, productive contributors, knowledge structure, and clusters since the 21st century.

METHODS

Documents were retrieved from the Web of Science Core Collection database from 2000 to 2024. VOSviewer, Incites, and Citespace were used to analyze the bibliometric metrics, which were categorized by country, institution, authors, journals, and keywords. The variables analyzed encompassed counts, citations, H-index, impact factor, and collaboration metrics.

RESULTS

Altogether, 7534 publications were initially retrieved and 2775 were included for analysis. The annual count and citation of papers exhibited exponential trends since 2018. The United States emerged as the lead contributor due to its high productivity and recognition levels. Stanford University, Johns Hopkins University, National University of Singapore, Mayo Clinic, University of Arizona, and University of Toronto were representative institutions in their respective fields. Cureus, JMIR Medical Education, Medical Teacher, and BMC Medical Education ranked as the top four most productive journals. The resulting heat map highlighted several high-frequency keywords, including performance, education, AI, and model. The citation burst time of terms revealed that AI technologies shifted from imaging processing (2000), augmented reality (2013), and virtual reality (2016) to decision-making (2020) and model (2021). Keywords such as mortality and robotic surgery persisted into 2023, suggesting the ongoing recognition and interest in these areas.

CONCLUSIONS

This study provides valuable insights and guidance for researchers who are interested in educational technology, as well as recommendations for pioneering institutions and journal submissions. Along with the rapid growth of AI, medical education is expected to gain much more benefits.

AI driven decision support reduces antibiotic mismatches and inappropriate use in outpatient urinary tract infections

Urinary tract infections (UTIs) often prompt empiric outpatient antibiotic prescriptions, risking mismatches. This study evaluates the impact of "UTI Smart-Set" (UTIS), an AI-driven decision-support tool, on prescribing patterns and mismatches in a large outpatient organization. UTIS integrates machine learning forecasts of antibiotic resistance, patient data, and guidelines into a user-friendly order set for UTI management. From 6/1/2021-8/31/2022, 171,010 UTI diagnoses were recorded, with UTIS used in 75,630 cases involving antibiotic prescriptions. Overall acceptance rate of UTIS recommendations was 66.0%. Among 19,287 cases with urine cultures, antibiotic mismatch rate was significantly lower when UTIS recommendations were followed (8.9% vs. 14.2%, p < 0.0001). Among women over 18, mismatch rate was 47.5% lower, and among women over 50, 55.6% lower (p < 0.001). Additionally, an overall reduction of 80.5% in ciprofloxacin usage (6.4% vs 32.9%, p < 0.0001) was observed. UTIS improved prescribing accuracy, reduced mismatches, and minimized quinolone use, highlighting AI's potential for personalized infection management.

Development and Validation of a Diagnostic Model for Stanford Type B Aortic Dissection Based on Proteomic Profiling

Purpose

Stanford Type B Aortic Dissection (TBAD), a critical aortic disease, has exhibited stable mortality rates over the past decade. However, diagnostic approaches for TBAD during routine health check-ups are currently lacking. This study focused on developing a model to improve the diagnosis in a population.

Patients and Methods

Serum biomarkers were investigated in 88 participants using proteomic profiling combined with machine learning. The findings were validated using ELISA in other 80 participants. Subsequently, a diagnostic model for TBAD integrating biomarkers with clinical indicators was developed and assessed using machine learning.

Results

Six differentially expressed proteins (DEPs) were identified through proteomic profiling and machine learning in discovery and derivation cohorts. Five of these (GDF-15, IL6, CD58, LY9, and Siglec-7) were further verified through ELISA validation within the validation cohort. In addition, ten blood-related indicators were selected as clinical indicators. Combining biomarkers and clinical indicators, the machine learning-based models performed well (AUC of the biomarker model = 0.865, AUC of the clinical model = 0.904, and AUC of the combined model = 0.909) using relative quantitation. The performance of the three models was verified (AUC of biomarker model = 0.866, AUC of clinical model = 0.868, and AUC of combined model = 0.886) using absolute quantitation. Crucially, the combined models outperformed individual biomarkers and clinical models, demonstrating superior efficacy.

Conclusion

Using proteomic profiling, we identified serum IL-6, GDF-15, CD58, LY9, and Siglec-7 as TBAD biomarkers. The machine-learning-based diagnostic model exhibited significant potential for TBAD diagnosis using only blood samples within the population.

Developing and validating a machine learning model to predict multidrug-resistant Klebsiella pneumoniae-related septic shock

Background

Multidrug-resistant Klebsiella pneumoniae (MDR-KP) infections pose a significant global healthcare challenge, particularly due to the high mortality risk associated with septic shock. This study aimed to develop and validate a machine learning-based model to predict the risk of MDR-KP-associated septic shock, enabling early risk stratification and targeted interventions.

Methods

A retrospective analysis was conducted on 1,385 patients with MDR-KP infections admitted between January 2019 and June 2024. The cohort was randomly divided into a training set (n = 969) and a validation set (n = 416). Feature selection was performed using LASSO regression and the Boruta algorithm. Seven machine learning algorithms were evaluated, with logistic regression chosen for its optimal balance between performance and robustness against overfitting.

Results

The overall incidence of MDR-KP-associated septic shock was 16.32% (226/1,385). The predictive model identified seven key risk factors: procalcitonin (PCT), sepsis, acute kidney injury, intra-abdominal infection, use of vasoactive medications, ventilator weaning failure, and mechanical ventilation. The logistic regression model demonstrated excellent predictive performance, with an area under the receiver operating characteristic curve (AUC) of 0.906 in the training set and 0.865 in the validation set. Calibration was robust, with Hosmer-Lemeshow test results of P = 0.065 (training) and P = 0.069 (validation). Decision curve analysis indicated substantial clinical net benefit.

Conclusion

This study presents a validated, high-performing predictive model for MDR-KP-associated septic shock, offering a valuable tool for early clinical decision-making. Prospective, multi-center studies are recommended to further evaluate its clinical applicability and effectiveness in diverse settings.

Harnessing artificial intelligence in sepsis care: advances in early detection, personalized treatment, and real-time monitoring

Sepsis remains a leading cause of morbidity and mortality worldwide due to its rapid progression and heterogeneous nature. This review explores the potential of Artificial Intelligence (AI) to transform sepsis management, from early detection to personalized treatment and real-time monitoring. AI, particularly through machine learning (ML) techniques such as random forest models and deep learning algorithms, has shown promise in analyzing electronic health record (EHR) data to identify patterns that enable early sepsis detection. For instance, random forest models have demonstrated high accuracy in predicting sepsis onset in intensive care unit (ICU) patients, while deep learning approaches have been applied to recognize complications such as sepsis-associated acute respiratory distress syndrome (ARDS). Personalized treatment plans developed through AI algorithms predict patient-specific responses to therapies, optimizing therapeutic efficacy and minimizing adverse effects. AI-driven continuous monitoring systems, including wearable devices, provide real-time predictions of sepsis-related complications, enabling timely interventions. Beyond these advancements, AI enhances diagnostic accuracy, predicts long-term outcomes, and supports dynamic risk assessment in clinical settings. However, ethical challenges, including data privacy concerns and algorithmic biases, must be addressed to ensure fair and effective implementation. The significance of this review lies in addressing the current limitations in sepsis management and highlighting how AI can overcome these hurdles. By leveraging AI, healthcare providers can significantly enhance diagnostic accuracy, optimize treatment protocols, and improve overall patient outcomes. Future research should focus on refining AI algorithms with diverse datasets, integrating emerging technologies, and fostering interdisciplinary collaboration to address these challenges and realize AI's transformative potential in sepsis care.

The role of artificial intelligence in pandemic responses: from epidemiological modeling to vaccine development

Integrating Artificial Intelligence (AI) across numerous disciplines has transformed the worldwide landscape of pandemic response. This review investigates the multidimensional role of AI in the pandemic, which arises as a global health crisis, and its role in preparedness and responses, ranging from enhanced epidemiological modelling to the acceleration of vaccine development. The confluence of AI technologies has guided us in a new era of data-driven decision-making, revolutionizing our ability to anticipate, mitigate, and treat infectious illnesses. The review begins by discussing the impact of a pandemic on emerging countries worldwide, elaborating on the critical significance of AI in epidemiological modelling, bringing data-driven decision-making, and enabling forecasting, mitigation and response to the pandemic. In epidemiology, AI-driven epidemiological models like SIR (Susceptible-Infectious-Recovered) and SIS (Susceptible-Infectious-Susceptible) are applied to predict the spread of disease, preventing outbreaks and optimising vaccine distribution. The review also demonstrates how Machine Learning (ML) algorithms and predictive analytics improve our knowledge of disease propagation patterns. The collaborative aspect of AI in vaccine discovery and clinical trials of various vaccines is emphasised, focusing on constructing AI-powered surveillance networks. Conclusively, the review presents a comprehensive assessment of how AI impacts epidemiological modelling, builds AI-enabled dynamic models by collaborating ML and Deep Learning (DL) techniques, and develops and implements vaccines and clinical trials. The review also focuses on screening, forecasting, contact tracing and monitoring the virus-causing pandemic. It advocates for sustained research, real-world implications, ethical application and strategic integration of AI technologies to strengthen our collective ability to face and alleviate the effects of global health issues.

Assessing the impact of antimicrobial resistance policies on antibiotic use and antimicrobial resistance-associated mortality in children and adults in low and middle-income countries: a global analysis

Introduction

Antimicrobial resistance (AMR) poses a major threat to global health security today. In recent years, many low and middle-income countries (LMICs) have implemented policies to optimise antibiotic use in both formal and informal healthcare settings. However, there is limited evidence on the effectiveness of these national efforts in LMICs.

Methods

We investigated the empirical relationship between national policies aimed at restricting antibiotic use and actual antibiotic consumption in 138 LMICs. Data on national policies were obtained from the Tripartite AMR Country Self-Assessment Survey (TrACSS) as well as from the Global Survey of Experts on AMR (GSEAR). Seven independent variables relating to AMR policies were evaluated. Outcomes included the proportion of children receiving antibiotics for lower respiratory tract infections and diarrhoea (specific to paediatric populations), along with total antibiotic consumption and AMR-associated mortality in general populations.

Results

Our analysis of 138 LMICs found wide variation in antibiotic use between countries and regions. We observed strong evidence of negative association (mean difference MD=-0.150, 95% CI (-0.2593 to -0.0407)) between the presence of regulatory or legislative policies that ban over-the-counter sales of antibiotics and the proportion of children receiving antibiotic drugs for lower respiratory tract infection. Furthermore, stronger AMR governance was associated with reduced total antibiotic consumption at the country level (MD=-1.259, 95% CI (-2.297 to -0.2216)). No associations were found between other policy variables and antibiotic use or mortality.

Conclusion

The results presented here suggest that there is some evidence of an empirical relationship between national policies aimed at limiting over-the-counter antibiotic sales and actual antibiotic usage practices. Further policy effectiveness research will be needed to better understand the true impact of government measures. In general, a multifaceted approach will likely be needed to fight AMR and preserve antibiotics' effectiveness, including evidence-based policies, targeted education and research.

A Systematic Review of Features Forecasting Patient Arrival Numbers

Adequate nurse staffing is crucial for quality healthcare, necessitating accurate predictions of patient arrival rates. These forecasts can be determined using supervised machine learning methods. Optimization of machine learning methods is largely about minimizing the prediction error. Existing models primarily utilize data such as historical patient visits, seasonal trends, holidays, and calendars. However, it is unclear what other features reduce the prediction error. Our systematic literature review identifies studies that use supervised machine learning to predict patient arrival numbers using nontemporal features, which are features not based on time or dates. We scrutinized 26 284 studies, eventually focusing on 27 relevant ones. These studies highlight three main feature groups: weather data, internet search and usage data, and data on (social) interaction of groups. Internet data and social interaction data appear particularly promising, with some studies reporting reduced errors by up to 33%. Although weather data are frequently used, its utility is less clear. Other potential data sources, including smartphone and social media data, remain largely unexplored. One reason for this might be potential data privacy challenges. In summary, although patient arrival prediction has become more important in recent years, there are still many questions and opportunities for future research on the features used in this area.

Development and Validation of a Machine Learning Model for the Prediction of Bloodstream Infections in Patients with Hematological Malignancies and Febrile Neutropenia

Background/Objectives: The rise of multidrug-resistant (MDR) infections demands personalized antibiotic strategies for febrile neutropenia (FN) in hematological malignancies. This study investigates machine learning (ML) for identifying patient profiles with increased susceptibility to bloodstream infections (BSI) during FN onset, aiming to tailor treatment approaches. Methods: From January 2020 to June 2022, we used the unsupervised ML algorithm KAMILA to analyze data from hospitalized hematological malignancy patients. Eleven features categorized clinical phenotypes and determined BSI and multidrug-resistant Gram-negative bacilli (MDR-GNB) prevalences at FN onset. Model performance was evaluated with a validation cohort from July 2022 to March 2023. Results: Among 462 FN episodes analyzed in the development cohort, 116 (25.1%) had BSIs. KAMILA's stratification identified three risk clusters: Cluster 1 (low risk), Cluster 2 (intermediate risk), and Cluster 3 (high risk). Cluster 2 (28.4% of episodes) and Cluster 3 (43.7%) exhibited higher BSI rates of 26.7% and 37.6% and GNB BSI rates of 13.4% and 19.3%, respectively. Cluster 3 had a higher incidence of MDR-GNB BSIs, accounting for 75% of all MDR-GNB BSIs. Cluster 1 (27.9% of episodes) showed a lower BSI risk (<1%) with no GNB infections. Validation cohort results were similar: Cluster 3 had a BSI rate of 38.1%, including 78% of all MDR-GNB BSIs, while Cluster 1 had no GNB-related BSIs. Conclusions: Unsupervised ML-based risk stratification enhances evidence-driven decision-making for empiric antibiotic therapies at FN onset, crucial in an era of rising multi-drug resistance.

Using Large Language Models to Retrieve Critical Data from Clinical Processes and Business Rules

Current clinical care relies heavily on complex, rule-based systems for tasks like diagnosis and treatment. However, these systems can be cumbersome and require constant updates. This study explores the potential of the large language model (LLM), LLaMA 2, to address these limitations. We tested LLaMA 2's performance in interpreting complex clinical process models, such as Mayo Clinic Care Pathway Models (CPMs), and providing accurate clinical recommendations. LLM was trained on encoded pathways versions using DOT language, embedding them with SentenceTransformer, and then presented with hypothetical patient cases. We compared the token-level accuracy between LLM output and the ground truth by measuring both node and edge accuracy. LLaMA 2 accurately retrieved the diagnosis, suggested further evaluation, and delivered appropriate management steps, all based on the pathways. The average node accuracy across the different pathways was 0.91 (SD ± 0.045), while the average edge accuracy was 0.92 (SD ± 0.122). This study highlights the potential of LLMs for healthcare information retrieval, especially when relevant data are provided. Future research should focus on improving these models' interpretability and their integration into existing clinical workflows.

Machine learning reveals CAT gene as a novel potential diagnostic and prognostic biomarker in non-small cell lung cancer

BACKGROUND

Non-small cell lung cancer (NSCLC) represents one of the most prevalent forms of lung cancer, with a five-year survival rate of 21.7%. There is an urgent need to identify pertinent biomarkers to inform the diagnosis and prognosis of tumors, particularly those that can be applied to different age groups. Herein, we would apply machine learning methods to specifically analyze the issue of biomarker applicability across different age groups in NSCLC.

METHODS

Studies have shown a higher incidence of NSCLC in people over 40 years of age, and due to the limitations of data set, studies of individuals under 40 years of age were not included in this study. To simulate the human aging model as closely as possible, we gathered corresponding non-small cell lung cancer (NSCLC) samples from the UCSC Xena database based on patient age information. These samples were then categorized into three groups: 40-60, 60-80, and over 80 years old. Subsequently, we employed four machine learning methods-Random Forest, LASSO regression analysis, XGBoost, and GBM-to identify gene sets with significant diagnostic value for each age group. By taking the intersection of these sets, we identified the optimal gene and assessed its prognostic significance in NSCLC. Then, the diagnostic value of CAT gene was validated using global public databases, including the GSE32863, GSE43458, GSE68571, GSE10072, and GSE63459 datasets from the Americas, the GSE30219 and GSE102511 datasets from Europe, and the GSE31210 and GSE19804 datasets from Asia. Furthermore, immunohistochemical staining was performed in an independent cohort from a tissue microarray. Additionally, cell culture and RT-qPCR were employed for external validation.

RESULTS

Through the implementation of machine learning methods, we successfully identified the catalase (CAT) gene. Our analysis revealed that individuals with high expression of the CAT gene experienced improved survival rates. Additionally, these individuals exhibited elevated immune scores. We further discovered that the CAT gene synergizes with multiple components of neutrophils, including TLRs, FcRn, and the selective GEF of Rho-family GTPases. In addition, we identified a potential immune checkpoint, TNFSF15, which is applicable to the human aging model. Finally, we validated the CAT gene's diagnostic value using databases encompassing the Americas, Europe, and Asia regions. Through external RT-qPCR validation, we verified that CAT expression in BEAS-2B was higher than that of A549. In an independent human cohort, we also verified that CAT is lowly expressed in lung cancer tissues. In addition, higher CAT levels were associated with improved survival in the 40-60 and 60-80 age groups.

CONCLUSIONS

In our analysis of the NSCLC database, we pinpointed the CAT gene, which holds promise for potential diagnostic and prognostic applications in the context of human aging. Furthermore, it may offer insights into addressing age-related heterogeneity of NSCLC.

Machine learning in infectious diseases: potential applications and limitations

Infectious diseases are a major threat for human and animal health worldwide. Artificial Intelligence (AI) combined algorithms including Machine Learning and Big Data analytics have emerged as a potential solution to analyse diverse datasets and face challenges posed by infectious diseases. In this commentary we explore the potential applications and limitations of ML to management of infectious disease. It explores challenges in key areas such as outbreak prediction, pathogen identification, drug discovery, and personalized medicine. We propose potential solutions to mitigate these hurdles and applications of ML to identify biomolecules for effective treatment and prevention of infectious diseases. In addition to use of ML for management of infectious diseases, potential applications are based on catastrophic evolution events for the identification of biomolecular targets to reduce risks for infectious diseases and vaccinomics for discovery and characterization of vaccine protective antigens using intelligent Big Data analytics techniques. These considerations set a foundation for developing effective strategies for managing infectious diseases in the future.

Using digital health technologies to optimise antimicrobial use globally

Digital health technology (DHT) describes tools and devices that generate or process health data. The application of DHTs could improve the diagnosis, treatment, and surveillance of bacterial infection and the prevention of antimicrobial resistance (AMR). DHTs to optimise antimicrobial use are rapidly being developed. To support the global adoption of DHTs and the opportunities offered to optimise antimicrobial use consensus is needed on what data are required to support antimicrobial decision making. This Series paper will explore bacterial AMR in humans and the need to optimise antimicrobial use in response to this global threat. It will also describe state-of-the-art DHTs to optimise antimicrobial prescribing in high-income and low-income and middle-income countries, and consider what fundamental data are ideally required for and from such technologies to support optimised antimicrobial use.

Comprehensive analysis of abnormal methylation modification differential expression mRNAs between low-grade and high-grade intervertebral disc degeneration and its correlation with immune cells

BACKGROUND

Intervertebral disc degeneration (IDD) is an important cause of low back pain. The aim of this study is to identify the potential molecular mechanism of abnormal methylation-modified DNA in the progression of IDD, hoping to contribute to the diagnosis and management of IDD.

METHODS

Low-grade IDD (grade I-II) and high-grade IDD (grade III-V) data were downloaded from GSE70362 and GSE129789 datasets. The abnormally methylated modified differentially expressed mRNAs (DEmRNAs) were identified by differential expression analysis (screening criteria were p < .05 and |logFC| > 1) and differential methylation analysis (screening criteria were p < .05 and |δβ| > 0.1). The classification models were constructed, and the receiver operating characteristic analysis was also carried out. In addition, functional enrichment analysis and immune correlation analysis were performed and the miRNAs targeted for the abnormally methylated DEmRNAs were predicted. Finally, expression validation was performed using real-time PCR.

RESULTS

Compared with low-grade IDD, seven abnormal methylation-modified DEmRNAs (AOX1, IBSP, QDPR, ABLIM1, CRISPLD2, ACTC1 and EMILIN1) were identified in high-grade IDD, and the classification models of random forests (RF) and support vector machine (SVM) were constructed. Moreover, seven abnormal methylation-modified DEmRNAs and classification models have high diagnostic accuracy (area under the curve [AUC] > 0.8). We also found that AUC values of single abnormal methylation-modified DEmRNA were all lower than those of RF and SVM classification models. Pearson correlation analysis found that macrophages M2 and EMILIN1 had significant negative correlation, while macrophages M2 and IBSP had significant positive correlation. In addition, four targeted relationship pairs (hsa-miR-4728-5p-QDPR, hsa-miR-4533-ABLIM1, hsa-miR-4728-5p-ABLIM1 and hsa-miR-4534-CRISPLD2) and multiple signalling pathways (for example, PI3K-AKT signalling pathway, osteoclast differentiation and calcium signalling pathway) were also identified that may be involved in the progression of IDD.

CONCLUSION

The identification of abnormal methylation-modified DEmRNAs and the construction of classification models in this study were helpful for the diagnosis and management of IDD progression.

Development of the Korean Quality Improvement Platform in Surgery (K-QIPS) program: a nationwide project to improve surgical quality and patient safety

PURPOSE

Improvements in surgical quality and patient safety are critical components of the healthcare system. Despite excellent cancer survival rates in Korea, there is a lack of standardized postoperative complication management systems. To address this gap, the Korean Surgical Society initiated the development of the Korean Quality Improvement Platform in Surgery (K-QIPS) program.

METHODS

K-QIPS was successfully launched in 87 general hospitals. This nationwide surgical quality improvement program covers 5 major surgical fields: gastric surgery, colorectal surgery, hepatectomy and liver transplantation, pancreatectomy, and kidney transplantation.

RESULTS

Common and surgery-specific complication platforms will be developed, and the program will work toward the implementation of an artificial intelligence-based complication prediction system and the provision of evidence-based feedback to participating institutions. K-QIPS represents a significant step toward improving surgical quality and patient safety in Korea.

CONCLUSION

This program aims to reduce postoperative complications, mortality, and medical costs by providing a standardized platform for complication management and prediction. The successful implementation of this nationwide project may provide a good model for other countries that are required to improve surgical outcomes and patient care.

Explore key genes of Crohn's disease based on glycerophospholipid metabolism: A comprehensive analysis Utilizing Mendelian Randomization, Multi-Omics integration, Machine Learning, and SHAP methodology

BACKGROUND AND AIMS

Crohn's disease (CD) is a chronic, complex inflammatory condition with increasing incidence and prevalence worldwide. However, the causes of CD remain incompletely understood. We identified CD-related metabolites, inflammatory factors, and key genes by Mendelian randomization (MR), multi-omics integration, machine learning (ML), and SHAP.

METHODS

We first performed a mediation MR analysis on 1400 serum metabolites, 91 inflammatory factors, and CD. We found that certain phospholipids are causally related to CD. In the scRNA-seq data, monocytes were categorized into high and low metabolism groups based on their glycerophospholipid metabolism scores. The differentially expressed genes of these two groups of cells were extracted, and transcription factor prediction, cell communication analysis, and GSEA analysis were performed. After further screening of differentially expressed genes (FDR<0.05, log2FC>1), least absolute shrinkage and selection operator (LASSO) regression was performed to obtain hub genes. Models for hub genes were built using the Catboost, XGboost, and NGboost methods. Further, we used the SHAP method to interpret the models and obtain the gene with the highest contribution to each model. Finally, qRT-PCR was used to verify the expression of these genes in the peripheral blood mononuclear cells (PBMC) of CD patients and healthy subjects.

RESULT

MR results showed 1-palmitoyl-2-stearoyl-gpc (16:0/18:0) levels, 1-stearoyl-2-arachidonoyl-GPI (18:0/20:4) levels, 1-arachidonoyl-gpc (20:4n6) levels, 1-palmitoyl-2-arachidonoyl-gpc (16:0/20:4n6) levels, and 1-arachidonoyl-GPE (20:4n6) levels were significantly associated with CD risk reduction (FDR<0.05), with CXCL9 acting as a mediation between these phospholipids and CD. The analysis identified 19 hub genes, with Catboost, XGboost, and NGboost achieving AUC of 0.91, 0.88, and 0.85, respectively. The SHAP methodology obtained the three genes with the highest model contribution: G0S2, S100A8, and PLAUR. The qRT-PCR results showed that the expression levels of S100A8 (p = 0.0003), G0S2 (p < 0.0001), and PLAUR (p = 0.0141) in the PBMC of CD patients were higher than healthy subjects.

CONCLUSION

MR findings suggest that certain phospholipids may lower CD risk. G0S2, S100A8, and PLAUR may be potential pathogenic genes in CD. These phospholipids and genes could serve as novel diagnostic and therapeutic targets for CD.

Predicting Treatment Outcomes in Patients with Drug-Resistant Tuberculosis and Human Immunodeficiency Virus Coinfection, Using Supervised Machine Learning Algorithm

Drug-resistant tuberculosis (DR-TB) and HIV coinfection present a conundrum to public health globally and the achievement of the global END TB strategy in 2035. A descriptive, retrospective review of medical records of patients, who were diagnosed with DR-TB and received treatment, was conducted. Student's t-test was performed to assess differences between two means and ANOVA between groups. The Chi-square test with or without trend or Fischer's exact test was used to test the degree of association of categorical variables. Logistic regression was used to determine predictors of DR-TB treatment outcomes. A decision tree classifier, which is a supervised machine learning algorithm, was also used. Python version 3.8. and R version 4.1.1 software were used for data analysis. A p-value of 0.05 with a 95% confidence interval (CI) was used to determine statistical significance. A total of 456 DR-TB patients were included in the study, with more male patients (n = 256, 56.1%) than female patients (n = 200, 43.9%). The overall treatment success rate was 61.4%. There was a significant decrease in the % of patients cured during the COVID-19 pandemic compared to the pre-pandemic period. Our findings showed that machine learning can be used to predict TB patients' treatment outcomes.

Advancing non-alcoholic fatty liver disease prediction: a comprehensive machine learning approach integrating SHAP interpretability and multi-cohort validation

Introduction

Non-alcoholic fatty liver disease (NAFLD) represents a major global health challenge, often undiagnosed because of suboptimal screening tools. Advances in machine learning (ML) offer potential improvements in predictive diagnostics, leveraging complex clinical datasets.

Methods

We utilized a comprehensive dataset from the Dryad database for model development and training and performed external validation using data from the National Health and Nutrition Examination Survey (NHANES) 2017-2020 cycles. Seven distinct ML models were developed and rigorously evaluated. Additionally, we employed the SHapley Additive exPlanations (SHAP) method to enhance the interpretability of the models, allowing for a detailed understanding of how each variable contributes to predictive outcomes.

Results

A total of 14,913 participants were eligible for this study. Among the seven constructed models, the light gradient boosting machine achieved the highest performance, with an area under the receiver operating characteristic curve of 0.90 in the internal validation set and 0.81 in the external NHANES validation cohort. In detailed performance metrics, it maintained an accuracy of 87%, a sensitivity of 92.9%, and an F1 score of 0.92. Key predictive variables identified included alanine aminotransferase, gammaglutamyl transpeptidase, triglyceride glucose-waist circumference, metabolic score for insulin resistance, and HbA1c, which are strongly associated with metabolic dysfunctions integral to NAFLD progression.

Conclusions

The integration of ML with SHAP interpretability provides a robust predictive tool for NAFLD, enhancing the early identification and potential management of the disease. The model's high accuracy and generalizability across diverse populations highlight its clinical utility, though future enhancements should include longitudinal data and lifestyle factors to refine risk assessments further.

AI Algorithms for Modeling the Risk, Progression, and Treatment of Sepsis, Including Early-Onset Sepsis-A Systematic Review

Sepsis remains a significant contributor to neonatal mortality worldwide. However, the nonspecific nature of sepsis symptoms in neonates often leads to the necessity of empirical treatment, placing a burden of ineffective treatment on patients. Furthermore, the global challenge of antimicrobial resistance is exacerbating the situation. Artificial intelligence (AI) is transforming medical practice and in hospital settings. AI shows great potential for assessing sepsis risk and devising optimal treatment strategies. Background/Objectives: This review aims to investigate the application of AI in the detection and management of neonatal sepsis. Methods: A systematic literature review (SLR) evaluating AI methods in modeling and classifying sepsis between 1 January 2014, and 1 January 2024, was conducted. PubMed, Scopus, Cochrane, and Web of Science were systematically searched for English-language studies focusing on neonatal sepsis. Results: The analyzed studies predominantly utilized retrospective electronic medical record (EMR) data to develop, validate, and test AI models to predict sepsis occurrence and relevant parameters. Key predictors included low gestational age, low birth weight, high results of C-reactive protein and white blood cell counts, and tachycardia and respiratory failure. Machine learning models such as logistic regression, random forest, K-nearest neighbor (KNN), support vector machine (SVM), and XGBoost demonstrated effectiveness in this context. Conclusions: The summarized results of this review highlight the great promise of AI as a clinical decision support system for diagnostics, risk assessment, and personalized therapy selection in managing neonatal sepsis.

Comparison of DNA methylation based classification models for precision diagnostics of central nervous system tumors

As part of the advancement in therapeutic decision-making for brain tumor patients at St. Jude Children's Research Hospital (SJCRH), we developed three robust classifiers, a deep learning neural network (NN), k-nearest neighbor (kNN), and random forest (RF), trained on a reference series DNA-methylation profiles to classify central nervous system (CNS) tumor types. The models' performance was rigorously validated against 2054 samples from two independent cohorts. In addition to classic metrics of model performance, we compared the robustness of the three models to reduced tumor purity, a critical consideration in the clinical utility of such classifiers. Our findings revealed that the NN model exhibited the highest accuracy and maintained a balance between precision and recall. The NN model was the most resistant to drops in performance associated with a reduction in tumor purity, showing good performance until the purity fell below 50%. Through rigorous validation, our study emphasizes the potential of DNA-methylation-based deep learning methods to improve precision medicine for brain tumor classification in the clinical setting.

Artificial intelligence-clinical decision support system for enhanced infectious disease management: Accelerating ceftazidime-avibactam resistance detection in Klebsiella pneumoniae

BACKGROUND

Effective and rapid diagnostic strategies are required to manage antibiotic resistance in Klebsiella pneumonia (KP). This study aimed to design an artificial intelligence-clinical decision support system (AI-CDSS) using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) and machine learning for the rapid detection of ceftazidime-avibactam (CZA) resistance in KP to improve clinical decision-making processes.

METHODS

Out of 107,721 bacterial samples, 675 specimens of KP with suspected multi-drug resistance were selected. These specimens were collected from a tertiary hospital and four secondary hospitals between 2022 and 2023 to evaluate CZA resistance. We used MALDI-TOF MS and machine learning to develop an AI-CDSS with enhanced speed of resistance detection.

RESULTS

Machine learning models, especially light gradient boosting machines (LGBM), exhibited an area under the curve (AUC) of 0.95, indicating high accuracy. The predictive models formed the core of our newly developed AI-CDSS, enabling clinical decisions quicker than traditional methods using culture and antibiotic susceptibility testing by a day.

CONCLUSIONS

The study confirms that MALDI-TOF MS, integrated with machine learning, can swiftly detect CZA resistance. Incorporating this insight into an AI-CDSS could transform clinical workflows, giving healthcare professionals immediate, crucial insights for shaping treatment plans. This approach promises to be a template for future anti-resistance strategies, emphasizing the vital importance of advanced diagnostics in enhancing public health outcomes.

Phenotypic Characterization of Intensive Care Patients With Infections: A Pilot Study of Host and Pathogen-Based Cluster Analysis

INTRODUCTION

Sepsis is a prevalent, albeit complex, disorder among critically ill patients and a "one-size-fits-all" approach does not seem applicable. Host intrinsic characteristics and microorganisms' particularities may influence response to therapy and outcomes. Attempting to group patients and microorganism characteristics may be an important step in developing and facilitating personalized infection treatment plans. This work intends to identify infected patients' clusters using clinical data that includes infection determinants: the isolated pathogen and the site of infection.

METHODS

In this retrospective analysis, we included patients with a microbiologically documented infection and non-infected controls. Patients admitted between January 2015 and December 2019 in the intensive care unit (ICU) were included (aged 17-95 years). Those with isolated microorganisms during their ICU stay were further analyzed using cluster analysis (hierarchical clustering and K-means; SPSS version 25.0). Four primary outcomes were addressed: ICU and hospital mortality rate and ICU and hospital length of stay (LOS).

RESULTS

This study included 1,923 patients, of whom 721 (37.5%) had at least one microbiological isolate during their ICU stay. Patients with at least one isolate identification were older (mean age 67.7 years vs. 65 years; p < 0.001) and had a higher ICU and hospital mortality (20.3% vs. 24.3%, p = 0.041; 26.9% vs. 38.4%, p < 0.001), as well as a longer LOS (median hospital LOS 8 vs. 18 days; p < 0.001) than patients without microorganisms identified. Patients with at least one isolated microorganism were divided into five different clusters. Notable differences were found in their ICU and hospital trajectories between clusters.

CONCLUSION

The cluster analysis approach provided valuable insights into the complex interplay between bacterial virulence, infection site, and patient outcomes in critical care medicine. Patients infected with bacteraemia by Gram-positive bacteria (cluster 2) or Enterobacteriaceae (Cluster 5) and fungal isolation in respiratory samples (Cluster 3) should prompt more aggressive clinical interventions, as these patients are more prone to die in the hospital.

An ultrasound-based ensemble machine learning model for the preoperative classification of pleomorphic adenoma and Warthin tumor in the parotid gland

OBJECTIVES

The preoperative classification of pleomorphic adenomas (PMA) and Warthin tumors (WT) in the parotid gland plays an essential role in determining therapeutic strategies. This study aims to develop and validate an ultrasound-based ensemble machine learning (USEML) model, employing nonradiative and noninvasive features to differentiate PMA from WT.

METHODS

A total of 203 patients with histologically confirmed PMA or WT who underwent parotidectomy from two centers were enrolled. Clinical factors, ultrasound (US) features, and radiomic features were extracted to develop three types of machine learning model: clinical models, US models, and USEML models. The diagnostic performance of the USEML model, as well as that of physicians based on experience, was evaluated and validated using receiver operating characteristic (ROC) curves in internal and external validation cohorts. DeLong's test was used for comparisons of AUCs. SHAP values were also utilized to explain the classification model.

RESULTS

The USEML model achieved the highest AUC of 0.891 (95% CI, 0.774-0.961), surpassing the AUCs of both the US (0.847; 95% CI, 0.720-0.932) and clinical (0.814; 95% CI, 0.682-0.908) models. The USEML model also outperformed physicians in both internal and external validation datasets (both p < 0.05). The sensitivity, specificity, negative predictive value, and positive predictive value of the USEML model and physician experience were 89.3%/75.0%, 87.5%/54.2%, 87.5%/65.6%, and 89.3%/65.0%, respectively.

CONCLUSIONS

The USEML model, incorporating clinical factors, ultrasound factors, and radiomic features, demonstrated efficient performance in distinguishing PMA from WT in the parotid gland.

CLINICAL RELEVANCE STATEMENT

This study developed a machine learning model for preoperative diagnosis of pleomorphic adenoma and Warthin tumor in the parotid gland based on clinical, ultrasound, and radiomic features. Furthermore, it outperformed physicians in an external validation dataset, indicating its potential for clinical application.

KEY POINTS

• Differentiating pleomorphic adenoma (PMA) and Warthin tumor (WT) affects management decisions and is currently done by invasive biopsy. • Integration of US-radiomic, clinical, and ultrasound findings in a machine learning model results in improved diagnostic accuracy. • The ultrasound-based ensemble machine learning (USEML) model consistently outperforms physicians, suggesting its potential applicability in clinical settings.

Artificial intelligence and prescription of antibiotic therapy: present and future

INTRODUCTION

In the past few years, the use of artificial intelligence in healthcare has grown exponentially. Prescription of antibiotics is not exempt from its rapid diffusion, and various machine learning (ML) techniques, from logistic regression to deep neural networks and large language models, have been explored in the literature to support decisions regarding antibiotic prescription.

AREAS COVERED

In this narrative review, we discuss promises and challenges of the application of ML-based clinical decision support systems (ML-CDSSs) for antibiotic prescription. A search was conducted in PubMed up to April 2024.

EXPERT OPINION

Prescribing antibiotics is a complex process involving various dynamic phases. In each of these phases, the support of ML-CDSSs has shown the potential, and also the actual ability in some studies, to favorably impacting relevant clinical outcomes. Nonetheless, before widely exploiting this massive potential, there are still crucial challenges ahead that are being intensively investigated, pertaining to the transparency of training data, the definition of the sufficient degree of prediction explanations when predictions are obtained through black box models, and the legal and ethical framework for decision responsibility whenever an antibiotic prescription is supported by ML-CDSSs.

Healthcare as a driver, reservoir and amplifier of antimicrobial resistance: opportunities for interventions

Antimicrobial resistance (AMR) is a global health challenge that threatens humans, animals and the environment. Evidence is emerging for a role of healthcare infrastructure, environments and patient pathways in promoting and maintaining AMR via direct and indirect mechanisms. Advances in vaccination and monoclonal antibody therapies together with integrated surveillance, rapid diagnostics, targeted antimicrobial therapy and infection control measures offer opportunities to address healthcare-associated AMR risks more effectively. Additionally, innovations in artificial intelligence, data linkage and intelligent systems can be used to better predict and reduce AMR and improve healthcare resilience. In this Review, we examine the mechanisms by which healthcare functions as a driver, reservoir and amplifier of AMR, contextualized within a One Health framework. We also explore the opportunities and innovative solutions that can be used to combat AMR throughout the patient journey. We provide a perspective on the current evidence for the effectiveness of interventions designed to mitigate healthcare-associated AMR and promote healthcare resilience within high-income and resource-limited settings, as well as the challenges associated with their implementation.

Proof of concept study on early forecasting of antimicrobial resistance in hospitalized patients using machine learning and simple bacterial ecology data

Antibiotic resistance in bacterial pathogens is a major threat to global health, exacerbated by the misuse of antibiotics. In hospital practice, results of bacterial cultures and antibiograms can take several days. Meanwhile, prescribing an empirical antimicrobial treatment is challenging, as clinicians must balance the antibiotic spectrum against the expected probability of susceptibility. We present here a proof of concept study of a machine learning-based system that predicts the probability of antimicrobial susceptibility and explains the contribution of the different cofactors in hospitalized patients, at four different stages prior to the antibiogram (sampling, direct examination, positive culture, and species identification), using only historical bacterial ecology data that can be easily collected from any laboratory information system (LIS) without GDPR restrictions once the data have been anonymised. A comparative analysis of different state-of-the-art machine learning and probabilistic methods was performed using 44,026 instances over 7 years from the Hôpital Européen Marseille, France. Our results show that multilayer dense neural networks and Bayesian models are suitable for early prediction of antibiotic susceptibility, with AUROCs reaching 0.88 at the positive culture stage and 0.92 at the species identification stage, and even 0.82 and 0.92, respectively, for the least frequent situations. Perspectives and potential clinical applications of the system are discussed.

Intestinal flora and inflammatory bowel disease: Causal relationships and predictive models

BACKGROUND

Inflammatory bowel disease (IBD), including Crohn's disease and ulcerative colitis, is significantly influenced by intestinal flora. Understanding the genetic and microbiotic interplay is crucial for IBD prediction and treatment.

METHODS

We used Mendelian randomization (MR), transcriptomic analysis, and machine learning techniques, integrating data from the MiBioGen Consortium and various GWAS datasets. SNPs associated with intestinal flora were mapped to genes, with LASSO regression refining gene selection. Differentially expressed genes (DEGs) and immune infiltration patterns were identified through transcriptomic analysis. Six machine learning models were used for predictive modeling.

FINDINGS

MR analysis identified 25 gut microbiota classifications causally related to IBD. SNP mapping and gene expression analysis highlighted 24 significant genes. Drug target MR and colocalization validated these genes' causal relationships with IBD. Key pathways identified included the PI3K-Akt signaling pathway and epithelial-mesenchymal transition. Immune infiltration analysis revealed distinct patterns between high and low LASSO score groups. Machine learning models demonstrated high predictive value, with soft voting enhancing reliability.

INTERPRETATION

By integrating MR, transcriptomic analysis, and sophisticated machine learning approaches, this study elucidates the causal relationships between intestinal flora and IBD. The application of machine learning not only enhanced predictive modeling but also offered new insights into IBD pathogenesis, highlighted potential therapeutic targets, and established a robust framework for predicting IBD onset.

Machine learning optimized DriverDetect software for high precision prediction of deleterious mutations in human cancers

The detection of cancer-driving mutations is important for understanding cancer pathology and therapeutics development. Prediction tools have been created to streamline the computation process. However, most tools available have heterogeneous sensitivity or specificity. We built a machine learning-derived algorithm, DriverDetect that combines the outputs of seven pre-existing tools to improve the prediction of candidate driver cancer mutations. The algorithm was trained with cancer gene-specific mutation datasets of cancer patients to identify cancer drivers. DriverDetect performed better than the individual tools or their combinations in the validation test. It has the potential to incorporate future novel prediction algorithms and can be retrained with new datasets, offering an expanded application to pan-cancer analysis for cross-cancer study. (115 words).

Artificial Intelligence to Close the Gap between Pharmacokinetic/Pharmacodynamic Targets and Clinical Outcomes in Critically Ill Patients: A Narrative Review on Beta Lactams

Antimicrobial dosing can be a complex challenge. Although a solid rationale exists for a link between antibiotic exposure and outcome, conflicting data suggest a poor correlation between pharmacokinetic/pharmacodynamic targets and infection control. Different reasons may lead to this discrepancy: poor tissue penetration by β-lactams due to inflammation and inadequate tissue perfusion; different bacterial response to antibiotics and biofilms; heterogeneity of the host's immune response and drug metabolism; bacterial tolerance and acquisition of resistance during therapy. Consequently, either a fixed dose of antibiotics or a fixed target concentration may be doomed to fail. The role of biomarkers in understanding and monitoring host response to infection is also incompletely defined. Nowadays, with the ever-growing stream of data collected in hospitals, utilizing the most efficient analytical tools may lead to better personalization of therapy. The rise of artificial intelligence and machine learning has allowed large amounts of data to be rapidly accessed and analyzed. These unsupervised learning models can apprehend the data structure and identify homogeneous subgroups, facilitating the individualization of medical interventions. This review aims to discuss the challenges of β-lactam dosing, focusing on its pharmacodynamics and the new challenges and opportunities arising from integrating machine learning algorithms to personalize patient treatment.

An Institutional Febrile Neutropenia Protocol Improved the Antibacterial Treatment and Encouraged the Development of a Computerized Clinical Decision Support System

We investigated the influence of a local guideline on the quality of febrile neutropenia (FN) management and the applicability of a computerized decision support system (CDSS) using real-life data. The study included 227 FN patients between April 2016 and January 2019. The primary outcome measure was the achievement of a 20% increase in the rate of appropriate empirical treatment of FN in bacteremic patients. The compatibility of the CDSS (the development of which was completed in November 2021) with local protocols was tested using standard patient scenarios and empirical antibiotic recommendations for bacteremic FN patients. In total, 91 patients were evaluated before (P1: between April 2016 and May 2017) and 136 after (P2: between May 2017 and January 2019) the guideline's release (May 2017). The demographic characteristics were similar. Appropriate empirical antibacterial treatment was achieved in 58.3% of P1 and 88.1% of P2 patients (p = 0.006). The need for escalation of antibacterial treatment was significantly lower in P2 (49.5% vs. 35.3%; p = 0.03). In P2, the performance of the CDSS and consulting physicians was similar (CDSS 88.8% vs. physician 88.83%; p = 1) regarding appropriate empirical antibacterial treatment. The introduction of the local guideline improved the appropriateness of initial empirical treatment and reduced escalation rates in FN patients. The high rate of compliance of the CDSS with the local guideline-based decisions in P2 highlights the usefulness of the CDSS for these patients.

Metabolism score and machine learning models for the prediction of esophageal squamous cell carcinoma progression

The incomplete prediction of prognosis in esophageal squamous cell carcinoma (ESCC) patients is attributed to various therapeutic interventions and complex prognostic factors. Consequently, there is a pressing demand for enhanced predictive biomarkers that can facilitate clinical management and treatment decisions. This study recruited 491 ESCC patients who underwent surgical treatment at Huashan Hospital, Fudan University. We incorporated 14 blood metabolic indicators and identified independent prognostic indicators for overall survival through univariate and multivariate analyses. Subsequently, a metabolism score formula was established based on the biochemical markers. We constructed a nomogram and machine learning models utilizing the metabolism score and clinically significant prognostic features, followed by an evaluation of their predictive accuracy and performance. We identified alkaline phosphatase, free fatty acids, homocysteine, lactate dehydrogenase, and triglycerides as independent prognostic indicators for ESCC. Subsequently, based on these five indicators, we established a metabolism score that serves as an independent prognostic factor in ESCC patients. By utilizing this metabolism score in conjunction with clinical features, a nomogram can precisely predict the prognosis of ESCC patients, achieving an area under the curve (AUC) of 0.89. The random forest (RF) model showed superior predictive ability (AUC = 0.90, accuracy = 86%, Matthews correlation coefficient = 0.55). Finally, we used an RF model with optimal performance to establish an online predictive tool. The metabolism score developed in this study serves as an independent prognostic indicator for ESCC patients.

Innovative strategies against superbugs: Developing an AI-CDSS for precise Stenotrophomonas maltophilia treatment

OBJECTIVES

The World Health Organization named Stenotrophomonas maltophilia (SM) a critical multi-drug resistant threat, necessitating rapid diagnostic strategies. Traditional culturing methods require up to 96 h, including 72 h for bacterial growth, identification with matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) through protein profile analysis, and 24 h for antibiotic susceptibility testing. In this study, we aimed at developing an artificial intelligence-clinical decision support system (AI-CDSS) by integrating MALDI-TOF MS and machine learning to quickly identify levofloxacin and trimethoprim/sulfamethoxazole resistance in SM, optimizing treatment decisions.

METHODS

We selected 8,662 SM from 165,299 MALDI-TOF MS-analysed bacterial specimens, collected from a major medical centre and four secondary hospitals. We exported mass-to-charge values and intensity spectral profiles from MALDI-TOF MS .mzML files to predict antibiotic susceptibility testing results, obtained with the VITEK-2 system using machine learning algorithms. We optimized the models with GridSearchCV and 5-fold cross-validation.

RESULTS

We identified distinct spectral differences between resistant and susceptible SM strains, demonstrating crucial resistance features. The machine learning models, including random forest, light-gradient boosting machine, and XGBoost, exhibited high accuracy. We established an AI-CDSS to offer healthcare professionals swift, data-driven advice on antibiotic use.

CONCLUSIONS

MALDI-TOF MS and machine learning integration into an AI-CDSS significantly improved rapid SM resistance detection. This system reduced the identification time of resistant strains from 24 h to minutes after MALDI-TOF MS identification, providing timely and data-driven guidance. Combining MALDI-TOF MS with machine learning could enhance clinical decision-making and improve SM infection treatment outcomes.

Identification of copper death-associated molecular clusters and immunological profiles for lumbar disc herniation based on the machine learning

Lumbar disc herniation (LDH) is a common clinical spinal disorder, yet its etiology remains unclear. We aimed to explore the role of cuproptosis-related genes (CRGs) and identify potential diagnostic biomarkers. Our analysis involved interrogating the GSE124272 and GSE150408 datasets for differential gene expression profiles associated with CRGs and immune characteristics. Molecular clustering was performed on LDH samples, followed by expression and immune infiltration analyses. Using the WGCNA algorithm, specific genes within CRG clusters were identified. After selecting the most predictive genes from the optimal model, four machine learning models were constructed and validated. This study identified nine CRGs associated with copper-regulated cell death. Two copper-containing molecular clusters linked to death were detected in LDH samples. Elevated expression and immune infiltration levels were found in LDH patients, particularly in CRG cluster C2. Utilizing XGB, five genes were identified for constructing a diagnostic model, achieving an area under the curve values of 0.715. In conclusion, this research provides valuable insights into the association between LDH and copper-regulated cell death, alongside proposing a promising predictive model.