Classification and Prediction of Sepsis Using Machine Learning and Deep Learning Techniques: A Comprehensive Literature Review

Abstract

Sepsis remains a leading cause of mortality worldwide, characterized by a dysregulated host response to infection that can rapidly progress to organ failure and death. Early detection and timely intervention are critical for improving patient outcomes, yet traditional clinical scoring systems often lack the sensitivity and specificity needed for optimal decision-making. This comprehensive literature review examines the state-of-the-art in sepsis classification and prediction using machine learning (ML) and deep learning (DL) techniques, based on an analysis of 235 recent research papers published from 2021 onwards. The review synthesizes findings on research methodologies, model architectures, commonly used datasets, and performance metrics. Key findings indicate that deep learning approaches, particularly Long Short-Term Memory (LSTM) networks and ensemble methods, consistently achieve superior predictive performance with AUROC values ranging from 0.85 to 0.99. The MIMIC-III and MIMIC-IV datasets emerge as the most widely used benchmarks, while the PhysioNet Challenge 2019 dataset provides standardized evaluation protocols. Despite promising results, challenges remain in reducing false alarm rates, ensuring model interpretability, and achieving generalization across diverse clinical settings. This review provides a comprehensive foundation for researchers and clinicians seeking to understand and implement AI-driven sepsis prediction systems.