Can Big Data and Machine Learning Improve Our Understanding of Acute Respiratory Distress Syndrome?

Abstract

Acute respiratory distress syndrome (ARDS) accounts for 10% of all diagnoses in the Intensive Care Unit, and about 40% of the patients succumb to the disease. Clinical methods alone can result in the under-recognition of this heterogeneous syndrome. The purpose of this study is to evaluate the role that big data and machine learning (ML) have played in understanding the heterogeneity of the disease and the development of various prediction algorithms. Most of the work in the field of ML in ARDS has been in the development of prediction models that have comparable efficacies to that of traditional models. Prediction algorithms have been useful in identifying new variables that may be important to consider in the future, supplementing the unknown information with the help of available noninvasive parameters, as well as predicting mortality. Phenotype identification using an unsupervised ML algorithm has been pivotal in classifying the heterogeneous population into more homogenous classes. Big data generated from ventilators in the form of ventilator waveform analysis and images in the form of radiomics have also been leveraged for the identification of the syndrome and can be incorporated into a clinical decision support system. Although the results are promising, lack of generalizability, “black box” nature of algorithms and concerns about “alarm fatigue” should be addressed for more mainstream adoption of these models.

Generating synthetic data with a mechanism-based Critical Illness Digital Twin: Demonstration for Post Traumatic Acute Respiratory Distress Syndrome

“…With respect to forecasting the potential progression to ARDS, numerous ML predictive algorithms that operate on data from the electronic health record [41, 42] (also see Ref [43] for a review) have been augmented with radiographic [44-46] and respiratory/ventilator waveform [47-49] data. As in other fields, such as precision oncology, there is also interest in being able to extend the feature set to include biomarkers and genetic/epigenetic information (see Ref [50] for a review). It is in this latter group that the ability to generate synthetic time series data becomes important using a mechanism-based Digital Twin.…”

Mechanical ventilation intervention based on machine learning from vital signs monitoring: a scoping review

“…Considering the MLA based supervised learning approach, a mapping via a mathematical function is used to correlate the inputs (predictors: MV parameters waveforms and/or VS monitoring data) and outputs (the outcome variables) [13]. The MLA based unsupervised learning approach could be used as a prior study to evaluate the cluster of health condition among the patients [43]. The predefined output MLA based unsupervised learning is not provided during the training process and is defined implicitly from the correlation among the various inputs.…”

Advances in Biomarkers for Diagnosis and Treatment of ARDS

“…Machine learning (ML) has presented evidence of its capacity to improve our understanding and management of ARDS. ML techniques have been implemented to predict the occurrence of ARDS, classify clinical phenotypes, and investigate the associations between biomarkers and ARDS outcomes [100][101][102][103]. The application and function of ML in the early diagnosis of ARDS are presented in Figure 2.…”