Machine learning predicts mortality based on analysis of ventilation parameters of critically ill patients: multi-centre validation

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

AIM

Acute respiratory distress syndrome or ARDS is an acute, severe form of respiratory failure characterised by poor oxygenation and bilateral pulmonary infiltrates. Advancements in signal processing and machine learning have led to promising solutions for classification, event detection and predictive models in the management of ARDS.

METHOD

In this review, we provide systematic description of different studies in the application of Machine Learning (ML) and artificial intelligence for management, prediction, and classification of ARDS. We searched the following databases: Google Scholar, PubMed, and EBSCO from 2009 to 2023. A total of 243 studies was screened, in which, 52 studies were included for review and analysis. We integrated knowledge of previous work providing the state of art and overview of explainable decision models in machine learning and have identified areas for future research.

RESULTS

Gradient boosting is the most common and successful method utilised in 12 (23.1%) of the studies. Due to limitation of data size available, neural network and its variation is used by only 8 (15.4%) studies. Whilst all studies used cross validating technique or separated database for validation, only 1 study validated the model with clinician input. Explainability methods were presented in 15 (28.8%) of studies with the most common method is feature importance which used 14 times.

CONCLUSION

For databases of 5000 or fewer samples, extreme gradient boosting has the highest probability of success. A large, multi-region, multi centre database is required to reduce bias and take advantage of neural network method. A framework for validating with and explaining ML model to clinicians involved in the management of ARDS would be very helpful for development and deployment of the ML model.

Advancing ICU patient care with a Real-Time predictive model for mechanical Power to mitigate VILI

BACKGROUND

Invasive Mechanical Ventilation (IMV) in Intensive Care Units (ICU) significantly increases the risk of Ventilator-Induced Lung Injury (VILI), necessitating careful management of mechanical power (MP). This study aims to develop a real-time predictive model of MP utilizing Artificial Intelligence to mitigate VILI.

METHODOLOGY

A retrospective observational study was conducted, extracting patient data from Clinical Information Systems from 2018 to 2022. Patients over 18 years old with more than 6 h of IMV were selected. Continuous data on IMV variables, laboratory data, monitoring, procedures, demographic data, type of admission, reason for admission, and APACHE II at admission were extracted. The variables with the highest correlation to MP were used for prediction and IMV data was grouped in 15-minute intervals using the mean. A mixed neural network model was developed to forecast MP 15 min in advance, using IMV data from 6 h before the prediction and current patient status. The model's ability to predict future MP was analyzed and compared to a baseline model predicting the future value of MP as equal to the current value.

RESULTS

The cohort consisted of 1967 patients after applying inclusion criteria, with a median age of 63 years and 66.9 % male. The deep learning model achieved a mean squared error of 2.79 in the test set, indicating a 20 % improvement over the baseline model. It demonstrated high accuracy (94 %) in predicting whether MP would exceed a critical threshold of 18 J/min, which correlates with increased mortality. The integration of this model into a web platform allows clinicians real-time access to MP predictions, facilitating timely adjustments to ventilation settings.

CONCLUSIONS

The study successfully developed and integrated in clinical practice a predictive model for MP. This model will assist clinicians allowing for the adjustment of ventilatory parameters before lung damage occurs.

Development and validation of an artificial intelligence mobile application for predicting 30-day mortality in critically ill patients with orthopaedic trauma

BACKGROUND

Given the intricate and grave nature of trauma-related injuries in ICU settings, it is imperative to develop and deploy reliable predictive tools that can aid in the early identification of high-risk patients who are at risk of early death. The objective of this study is to create and validate an artificial intelligence (AI) model that can accurately predict early mortality among critical fracture patients.

METHODS

A total of 2662 critically ill patients with orthopaedic trauma were included from the MIMIC III database. Early mortality was defined as death within 30 days in this study. The patients were randomly divided into a model training cohort and a model validation cohort. Various algorithms, including logistic regression (LR), extreme gradient boosting machine (eXGBM), decision tree (DT), support vector machine (SVM), random forest (RF), and neural network (NN), were employed. Evaluation metrics, including discrimination and calibration, were used to develop a comprehensive scoring system ranging from 0 to 60, with higher scores indicating better prediction performance. Furthermore, external validation was carried out using 131 patients. The optimal model was deployed as an internet-based AI tool.

RESULTS

Among all models, the eXGBM demonstrated the highest area under the curve (AUC) value (0.974, 95%CI: 0.959-0.983), followed by the RF model (0.951, 95%CI: 0.935-0.967) and the NN model (0.922, 95%CI: 0.905-0.941). Additionally, the eXGBM model outperformed other models in terms of accuracy (0.915), precision (0.906), recall (0.926), F1 score (0.916), Brier score (0.062), log loss (0.210), and discrimination slope (0.767). Based on the scoring system, the eXGBM model achieved the highest score (53), followed by RF (42) and NN (39). The LR, DT, and SVM models obtained scores of 28, 18, and 32, respectively. Decision curve analysis further confirmed the superior clinical net benefits of the eXGBM model. External validation of the model achieved an AUC value of 0.913 (95%CI: 0.878-0.948). Consequently, the model was deployed on the Internet at https://30-daymortalityincriticallyillpatients-fnfsynbpbp6rgineaspuim.streamlit.app/, allowing users to input patient features and obtain predicted risks of early mortality among critical fracture patients. Furthermore, the AI model successfully stratified patients into low or high risk of early mortality based on a predefined threshold and provided recommendations for appropriate therapeutic interventions.

CONCLUSION

This study successfully develops and validates an AI model, with the eXGBM algorithm demonstrating the highest predictive performance for early mortality in critical fracture patients. By deploying the model as a web-based AI application, healthcare professionals can easily access the tool, enabling them to predict 30-day mortality and aiding in the identification and management of high-risk patients among those critically ill with orthopedic trauma.

Using time-course as an essential factor to accurately predict sepsis-associated mortality among patients with suspected sepsis

BACKGROUND

Biomarker dynamics in different time-courses might be the primary reason why a static measurement of a single biomarker cannot accurately predict sepsis outcomes. Therefore, we conducted this prospective hospital-based cohort study to simultaneously evaluate the performance of several conventional and novel biomarkers of sepsis in predicting sepsis-associated mortality on different days of illness among patients with suspected sepsis.

METHODS

We evaluated the performance of 15 novel biomarkers including angiopoietin-2, pentraxin 3, sTREM-1, ICAM-1, VCAM-1, sCD14 and 163, E-selectin, P-selectin, TNF-alpha, interferon-gamma, CD64, IL-6, 8, and 10, along with few conventional markers for predicting sepsis-associated mortality. Patients were grouped into quartiles according to the number of days since symptom onset. Receiver operating characteristic curve (ROC) analysis was used to evaluate the biomarker performance.

RESULTS

From 2014 to 2017, 1483 patients were enrolled, of which 78% fulfilled the systemic inflammatory response syndrome criteria, 62% fulfilled the sepsis-3 criteria, 32% had septic shock, and 3.3% developed sepsis-associated mortality. IL-6, pentraxin 3, sCD163, and the blood gas profile demonstrated better performance in the early days of illness, both before and after adjusting for potential confounders (adjusted area under ROC curve [AUROC]:0.81-0.88). Notably, the Sequential Organ Failure Assessment (SOFA) score was relatively consistent throughout the course of illness (adjusted AUROC:0.70-0.91).

CONCLUSION

IL-6, pentraxin 3, sCD163, and the blood gas profile showed excellent predictive accuracy in the early days of illness. The SOFA score was consistently predictive of sepsis-associated mortality throughout the course of illness, with an acceptable performance.

Machine learning predicts lung recruitment in acute respiratory distress syndrome using single lung CT scan

BACKGROUND

To develop and validate classifier models that could be used to identify patients with a high percentage of potentially recruitable lung from readily available clinical data and from single CT scan quantitative analysis at intensive care unit admission. 221 retrospectively enrolled mechanically ventilated, sedated and paralyzed patients with acute respiratory distress syndrome (ARDS) underwent a PEEP trial at 5 and 15 cmH2O of PEEP and two lung CT scans performed at 5 and 45 cmH2O of airway pressure. Lung recruitability was defined at first as percent change in not aerated tissue between 5 and 45 cmH2O (radiologically defined; recruiters: Δ45-5non-aerated tissue  > 15%) and secondly as change in PaO2 between 5 and 15 cmH2O (gas exchange-defined; recruiters: Δ15-5PaO2  > 24 mmHg). Four machine learning (ML) algorithms were evaluated as classifiers of radiologically defined and gas exchange-defined lung recruiters using different models including different variables, separately or combined, of lung mechanics, gas exchange and CT data.

RESULTS

ML algorithms based on CT scan data at 5 cmH2O classified radiologically defined lung recruiters with similar AUC as ML based on the combination of lung mechanics, gas exchange and CT data. ML algorithm based on CT scan data classified gas exchange-defined lung recruiters with the highest AUC.

CONCLUSIONS

ML based on a single CT data at 5 cmH2O represented an easy-to-apply tool to classify ARDS patients in recruiters and non-recruiters according to both radiologically defined and gas exchange-defined lung recruitment within the first 48 h from the start of mechanical ventilation.

A Survey on Medical Explainable AI (XAI): Recent Progress, Explainability Approach, Human Interaction and Scoring System

The emerging field of eXplainable AI (XAI) in the medical domain is considered to be of utmost importance. Meanwhile, incorporating explanations in the medical domain with respect to legal and ethical AI is necessary to understand detailed decisions, results, and current status of the patient's conditions. Successively, we will be presenting a detailed survey for the medical XAI with the model enhancements, evaluation methods, significant overview of case studies with open box architecture, medical open datasets, and future improvements. Potential differences in AI and XAI methods are provided with the recent XAI methods stated as (i) local and global methods for preprocessing, (ii) knowledge base and distillation algorithms, and (iii) interpretable machine learning. XAI characteristics details with future healthcare explainability is included prominently, whereas the pre-requisite provides insights for the brainstorming sessions before beginning a medical XAI project. Practical case study determines the recent XAI progress leading to the advance developments within the medical field. Ultimately, this survey proposes critical ideas surrounding a user-in-the-loop approach, with an emphasis on human-machine collaboration, to better produce explainable solutions. The surrounding details of the XAI feedback system for human rating-based evaluation provides intelligible insights into a constructive method to produce human enforced explanation feedback. For a long time, XAI limitations of the ratings, scores and grading are present. Therefore, a novel XAI recommendation system and XAI scoring system are designed and approached from this work. Additionally, this paper encourages the importance of implementing explainable solutions into the high impact medical field.

Management of intoxicated patients – a descriptive outcome analysis of 4,267 ICU patients

Introduction

Intoxications are common in intensive care units (ICUs). The number of causative substances is large, mortality usually low. This retrospective cohort study aims to characterize differences of intoxicated compared to general ICU patients, point out variations according to causative agents, as well as to highlight differences between survivors and non-survivors among intoxicated individuals in a large-scale multi-center analysis.

Methods

A total of 105,998 general ICU patients and 4,267 individuals with the admission diagnoses “overdose” and “drug toxicity” from the years 2014 and 2015 where included from the eICU Collaborative Research Database. In addition to comparing these groups with respect to baseline characteristics, intensive care measures and outcome parameters, differences between survivors and non-survivors from the intoxication group, as well as the individual groups of causative substances were investigated.

Results

Intoxicated patients were younger (median 41 vs. 66 years; p<0.001), more often female (55 vs. 45%; p<0.001), and normal weighted (36% vs. 30%; p<0.001), whereas more obese individuals where observed in the other group (37 vs. 31%; p<0.001). Intoxicated individuals had a significantly lower mortality compared to general ICU patients (1% vs. 10%; aOR 0.07 95%CI 0.05-0.11; p<0.001), a finding which persisted after multivariable adjustment (aOR 0.17 95%CI 0.12-0.24; p<0.001) and persisted in all subgroups. Markers of disease severity (SOFA-score: 3 (1-5) vs. 4 (2-6) pts.; p<0.001) and frequency of vasopressor use (5 vs. 15%; p<0.001) where lower, whereas rates of mechanical ventilation where higher (24 vs. 26%; p<0.001) in intoxicated individuals. There were no differences with regard to renal replacement therapy in the first three days (3 vs. 4%; p=0.26). In sensitivity analysis (interactions for age, sex, ethnicity, hospital category, maximum initial lactate, mechanical ventilation, and vasopressor use), a trend towards lower mortality in intoxicated patients persisted in all subgroups.

Conclusion

This large-scale retrospective analysis indicates a significantly lower mortality of intoxicated individuals compared to general ICU patients.

Multilayer perceptron neural network model development for mechanical ventilator parameters prediction by real time system learning

Background and objective

In pandemic situations like COVID 19, real time monitoring of patient condition and continuous delivery of inspired oxygen can be made possible only through artificial intelligence-based system modeling. Even now manual control of mechanical ventilator parameters is continuing despite the ever-increasing number of patients in critical epidemic conditions. Here a suggestive multi-layer perceptron neural network model is developed to predict the level of inspired oxygen delivered by the mechanical ventilator along with mode and positive end expiratory pressure (PEEP) changes for reducing the effort of health care professionals.

Methods

The artificial neural network model is developed by Python programming using real time data. Parameter identification for model inputs and outputs is done by in corporating consistent real time patient data including periodical arterial blood gas analysis, continuous pulse oximetry readings and mechanical ventilator settings using statistical pairwise analysis using R programming.

Results

Mean square error values and R values of the model are calculated and found to be an average of 0.093 and 0.81 respectively for various data sets. Accuracy loss will be in good fit with validation loss for a comparable number of epochs.

Conclusions

Comparison of the model output is undertaken with physician’s prediction using statistical analysis and shows an accuracy error of 4.11 percentages which is permissible for a good predictive system.

Personalized mechanical ventilation in acute respiratory distress syndrome

A personalized mechanical ventilation approach for patients with adult respiratory distress syndrome (ARDS) based on lung physiology and morphology, ARDS etiology, lung imaging, and biological phenotypes may improve ventilation practice and outcome. However, additional research is warranted before personalized mechanical ventilation strategies can be applied at the bedside. Ventilatory parameters should be titrated based on close monitoring of targeted physiologic variables and individualized goals. Although low tidal volume (VT) is a standard of care, further individualization of VT may necessitate the evaluation of lung volume reserve (e.g., inspiratory capacity). Low driving pressures provide a target for clinicians to adjust VT and possibly to optimize positive end-expiratory pressure (PEEP), while maintaining plateau pressures below safety thresholds. Esophageal pressure monitoring allows estimation of transpulmonary pressure, but its use requires technical skill and correct physiologic interpretation for clinical application at the bedside. Mechanical power considers ventilatory parameters as a whole in the optimization of ventilation setting, but further studies are necessary to assess its clinical relevance. The identification of recruitability in patients with ARDS is essential to titrate and individualize PEEP. To define gas-exchange targets for individual patients, clinicians should consider issues related to oxygen transport and dead space. In this review, we discuss the rationale for personalized approaches to mechanical ventilation for patients with ARDS, the role of lung imaging, phenotype identification, physiologically based individualized approaches to ventilation, and a future research agenda.