Forecasting adverse surgical events using self-supervised transfer learning for physiological signals

TL;DR

This paper introduces PHASE, a self-supervised transfer learning method for physiological signals that improves the prediction of adverse surgical events across datasets, offering higher accuracy and explainability.

Contribution

The paper presents PHASE, a novel transferable embedding technique for physiological signals that enhances predictive accuracy and interpretability in surgical outcome forecasting.

Findings

PHASE outperforms state-of-the-art models in predicting adverse events.

Transfer learning with PHASE improves accuracy on unseen data.

PHASE is computationally efficient and explainable.

Abstract

Hundreds of millions of surgical procedures take place annually across the world, which generate a prevalent type of electronic health record (EHR) data comprising time series physiological signals. Here, we present a transferable embedding method (i.e., a method to transform time series signals into input features for predictive machine learning models) named PHASE (PHysiologicAl Signal Embeddings) that enables us to more accurately forecast adverse surgical outcomes based on physiological signals. We evaluate PHASE on minute-by-minute EHR data of more than 50,000 surgeries from two operating room (OR) datasets and patient stays in an intensive care unit (ICU) dataset. PHASE outperforms other state-of-the-art approaches, such as long-short term memory networks trained on raw data and gradient boosted trees trained on handcrafted features, in predicting five distinct outcomes: hypoxemia, hypocapnia, hypotension, hypertension, and phenylephrine administration. In a transfer learning setting where we train embedding models in one dataset then embed signals and predict adverse events in unseen data, PHASE achieves significantly higher prediction accuracy at lower computational cost compared to conventional approaches. Finally, given the importance of understanding models in clinical applications we demonstrate that PHASE is explainable and validate our predictive models using local feature attribution methods.