Detection of sepsis using biomarkers based on machine learning.

Area of Science:

• Critical Care Medicine
• Biomedical Engineering
• Clinical Diagnostics

Background:

• Sepsis is a leading cause of mortality in intensive care units (ICUs), particularly among patients with non-cardiovascular diseases.
• It ranks among the top ten causes of death for all hospitalized patients.
• Effective diagnostic methods are crucial for timely intervention and improved patient outcomes.

Purpose of the Study:

• To compare the diagnostic value of specific blood parameters for sepsis.
• To investigate the relationship between physiological parameters and sepsis.
• To identify a more practical diagnostic method for sepsis in the ICU.

Main Methods:

• A descriptive-analytical study involving 208 sepsis patients in the ICU.
• Measurement of physiological parameters (HR, O2Sat, SBP) in patients and healthy individuals.
• Data analysis using statistical methods and MATLAB, with classification via MLP, RBF, and KNN algorithms.

Main Results:

• Significant differences observed in HR, O2Sat, and SBP between sepsis patients and normal individuals.
• Classification models achieved high performance: 98% specificity, 100% sensitivity, and 99% accuracy.
• Machine learning classifiers (MLP, RBF, KNN) demonstrated robust diagnostic capabilities.

Conclusions:

• Physiological parameters (HR, O2Sat, SBP) show significant variations in sepsis patients.
• These simple tests could serve as practical alternatives to C3 and PCT for sepsis diagnosis in ICUs.
• Accurate and early sepsis detection is vital for enhancing patient treatment and prognosis.