Revolutionizing sepsis diagnosis using machine learning and deep learning models: a systematic literature review

Area of Science:

• Medical Informatics
• Computational Biology
• Clinical Data Science

Background:

• Sepsis is a critical condition requiring rapid diagnosis and treatment to prevent organ failure and death.
• Early detection of sepsis is crucial for improving patient outcomes and survival rates.
• Artificial intelligence (AI), Machine Learning (ML), and Deep Learning (DL) offer potential for enhanced early sepsis detection.

Purpose of the Study:

• To systematically review and synthesize existing ML/DL approaches for sepsis prediction, focusing on intensive care unit (ICU) settings.
• To analyze diverse data sources, feature selection, algorithms, preprocessing techniques, and evaluation metrics used in ML/DL sepsis prediction.
• To provide a comparative visual of diagnostic performance using a forest plot of AUC and sensitivity values.

Main Methods:

• Systematic review of 80 studies on ML/DL for sepsis prediction.
• Analysis of various data sources (e.g., MIMIC-III, eICU), algorithms (e.g., logistic regression, LSTM, transformers), and preprocessing techniques.
• Inclusion of a forest plot to visually summarize diagnostic performance metrics (AUC, sensitivity).

Main Results:

• A wide range of ML/DL models, from traditional to advanced deep learning architectures, have been applied to sepsis prediction.
• The review highlights variability in model performance and generalizability across different clinical datasets.
• Key challenges identified include model interpretability, ethical considerations, and insufficient external/temporal validation.

Conclusions:

• ML/DL models hold significant potential for improving early sepsis detection and patient care.
• Future research should focus on enhancing model interpretability, ensuring robust validation, and leveraging real-time data for clinical deployment.
• Actionable recommendations are provided for developing more robust, interpretable, and clinically relevant AI models for sepsis management.