Sepsis is a life-threatening medical condition caused by the body's extreme response to infection, which can lead to organ failure and death if not detected and treated early. Timely identification of sepsis remains a significant challenge in healthcare systems worldwide due to the complex physiological patterns and non-specific early symptoms. This paper presents a comprehensive machine learning and deep learning-based approach for early prediction of sepsis using patient vital signs and physiological parameters. The proposed system integrates multiple predictive models including Logistic Regression, Random Forest, XGBoost, and Long Short-Term Memory (LSTM) neural networks to enhance prediction accuracy and reliability. The methodology incorporates extensive data preprocessing techniques including median imputation for missing values, z-score normalization for feature scaling, and correlation-based feature selection. The system processes critical clinical parameters including heart rate, body temperature, systolic blood pressure, mean arterial pressure, oxygen saturation, respiratory rate, and various laboratory values. Experimental evaluation is conducted using a comprehensive dataset of ICU patient records with rigorous performance assessment through accuracy, precision, recall, F1-score, and Area Under the Receiver Operating Characteristic Curve (AUC-ROC). Results demonstrate that ensemble methods, particularly XGBoost, achieve superior discriminative performance with an accuracy of 98.2\% and the highest AUC-ROC of 0.7114 among all evaluated models. The integration of risk stratification capabilities enables clinical decision support through three-tier classification (low, medium, high risk), facilitating personalized patient management and resource optimization in critical care settings.
K.V.Kiran et al. (Thu,) studied this question.