,
Bharath Institute of Higher Education and Research , Chennai , India
Bharath Institute of Higher Education and Research , Chennai , India
Arterial Blood Gas (ABG) analysis is an important diagnostic tool in intensive care unit (ICU) settings that provides valuable information about the patient's respiratory and metabolic status. However, in the absence of predictive information, when testing is over-utilized or not planned, it can cause discomfort to the patient, costs to the healthcare system, and overtax already burdened resources. This work develops a predictive model that employs machine learning to classify acid-base imbalances and guides testing, which will advance diagnosis and efficiency in practice. The primary data source for model development was a dataset that included ABG profiles of ICU patients along with parameters of pH, PaCO₂, HCO₃⁻, PaO₂, lactate, and clinical indications of hemodynamic stability, respiratory support, and therapeutic interventions. Data pre-processing included: normalization, missing value imputation, and feature scaling, and the Synthetic Minority Over-sampling Technique (SMOTE) was used to create better class balance to improve generalization. The predictive utility used a family of Support Vector Machine (SVM) classifiers with linear, polynomial, and radial basis function (RBF) kernels, which were tuned using a grid search and 10-fold cross-validation. The implementation framework was created in Python 3.11 using Scikit-learn, NumPy, and Panda’s libraries. The optimized SVM classifier achieved a maximum accuracy of 93.02%, F-measure of 92.8%, precision of 93%, and an area under the ROC curve (AUC) of 0.97, for test data. The incorporation of SMOTE resulted in better class balance. This is the first application of its kind, exploring machine learning algorithms to achieve such high-performance metrics in the analysis of clinical ABG data obtained in the ICU, supporting and enhancing healthcare diagnostics.
This is an open access article distributed under the Creative Commons Attribution Non-Commercial License (CC BY-NC) License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
0
The statements, opinions and data contained in the journal are solely those of the individual authors and contributors and not of the publisher and the editor(s). We stay neutral with regard to jurisdictional claims in published maps and institutional affiliations.