Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Video

Updated: Oct 10, 2025

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Published on: October 10, 2018

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

Hugh Chen¹, Scott M Lundberg², Gabriel Erion^1,3

¹Paul G. Allen School of Computer Science and Engineering, University of Washington, 185 E Stevens Way NE, Seattle, WA, 98195, USA.

NPJ Digital Medicine

|December 9, 2021

Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Dissecting and directing pathology foundation models.

bioRxiv : the preprint server for biology·2026

Same author

DREAM: A framework for discovering mechanisms underlying AI prediction of protected attributes.

medRxiv : the preprint server for health sciences·2025

Same author

Deep profiling of gene expression across 18 human cancers.

Nature biomedical engineering·2024

Same author

Transparent medical image AI via an image-text foundation model grounded in medical literature.

Nature medicine·2024

Same author

Auditing the inference processes of medical-image classifiers by leveraging generative AI and the expertise of physicians.

Nature biomedical engineering·2023

Same author

Isolating salient variations of interest in single-cell data with contrastiveVI.

Nature methods·2023

We developed PHASE, a new method for analyzing physiological signals from electronic health records to predict adverse surgical outcomes. PHASE improves accuracy and efficiency in forecasting events like hypoxemia and hypotension.

Area of Science:

Biomedical Informatics
Machine Learning in Healthcare
Critical Care Medicine

Background:

Electronic health records (EHRs) contain valuable time series physiological signals from millions of surgeries.
Predicting adverse surgical outcomes from these signals is crucial for patient safety.
Current methods often struggle with the complexity and volume of physiological data.

Purpose of the Study:

To introduce PHASE (PHysiologicAl Signal Embeddings), a novel transferable embedding method for physiological signals.
To enhance the accuracy of predicting adverse surgical outcomes using machine learning.
To demonstrate the explainability and clinical applicability of the developed model.

Main Methods:

Developed PHASE, a transferable embedding technique to convert time series physiological signals into predictive features.

Related Experiment Videos

Last Updated: Oct 10, 2025

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Published on: October 10, 2018

Evaluated PHASE on minute-by-minute EHR data from over 50,000 surgeries across operating room (OR) and intensive care unit (ICU) datasets.

Compared PHASE against state-of-the-art methods including LSTM networks and gradient boosted trees.

Main Results:

PHASE significantly outperformed existing methods in predicting six adverse outcomes: hypoxemia, hypocapnia, hypotension, hypertension, phenylephrine, and epinephrine.
In transfer learning scenarios, PHASE achieved higher prediction accuracy with lower computational cost on unseen data.
The predictive models built with PHASE were validated as explainable using local feature attribution.

Conclusions:

PHASE offers a more accurate and computationally efficient approach for predicting adverse surgical events from physiological signals.
The method's transfer learning capability and explainability support its potential for clinical integration.
This work advances the use of machine learning for real-time patient monitoring and outcome prediction in surgical settings.