End to End Autoencoder MLP Framework for Sepsis Prediction

TL;DR

This paper presents an end-to-end deep learning framework combining an autoencoder and MLP for early sepsis prediction, outperforming traditional methods in accuracy and robustness across multiple ICU datasets.

Contribution

It introduces a novel integrated autoencoder-MLP framework with customized sampling and real-time inference strategies for improved sepsis prediction.

Findings

Achieved up to 93.5% accuracy in sepsis prediction.

Outperformed traditional machine learning models in multiple ICU cohorts.

Demonstrated robustness and generalizability across heterogeneous datasets.

Abstract

Sepsis is a life threatening condition that requires timely detection in intensive care settings. Traditional machine learning approaches, including Naive Bayes, Support Vector Machine (SVM), Random Forest, and XGBoost, often rely on manual feature engineering and struggle with irregular, incomplete time-series data commonly present in electronic health records. We introduce an end-to-end deep learning framework integrating an unsupervised autoencoder for automatic feature extraction with a multilayer perceptron classifier for binary sepsis risk prediction. To enhance clinical applicability, we implement a customized down sampling strategy that extracts high information density segments during training and a non-overlapping dynamic sliding window mechanism for real-time inference. Preprocessed time series data are represented as fixed dimension vectors with explicit missingness indicators, mitigating bias and noise. We validate our approach on three ICU cohorts. Our end-to-end model achieves accuracies of 74.6 percent, 80.6 percent, and 93.5 percent, respectively, consistently outperforming traditional machine learning baselines. These results demonstrate the framework's superior robustness, generalizability, and clinical utility for early sepsis detection across heterogeneous ICU environments.