Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Interpretable Deep Models for ICU Outcome Prediction.

Zhengping Che1, Sanjay Purushotham1, Robinder Khemani2

  • 1University of Southern California, Los Angeles, CA, USA.

AMIA ... Annual Symposium Proceedings. AMIA Symposium
|March 9, 2017
PubMed
Summary
This summary is machine-generated.

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

Development of a neurologic deterioration risk score in pediatric mild traumatic brain injury and intracranial injuries.

The journal of trauma and acute care surgery·2026
Same author

Machine-learning detection and variability of mesospheric frontal waves observed by VIIRS day/night band.

Earth, planets, and space : EPS·2025
Same author

A Master Protocol Template for Pediatric ARDS Studies.

Respiratory care·2024
Same author

Association of Interventions With Outcomes in Children At-Risk for Pediatric Acute Respiratory Distress Syndrome: A Pediatric Acute Respiratory Distress Syndrome Incidence and Epidemiology Study.

Pediatric critical care medicine : a journal of the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies·2023
Same author

Natural sub-bituminous coal as filler enhances mechanical, insulation and flame retardant properties of coir-polypropylene bio-composites.

Environmental geochemistry and health·2023
Same author

Methodology of the Second Pediatric Acute Lung Injury Consensus Conference.

Pediatric critical care medicine : a journal of the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies·2023
Same journal

Sensitivity Analyses of a Scoring System for a Contraception Decision Aid.

AMIA ... Annual Symposium proceedings. AMIA Symposium·2026
Same journal

Improving electronic health record processing of large language models via retrieval-augmented generation: A case study on dietary supplements.

AMIA ... Annual Symposium proceedings. AMIA Symposium·2026
Same journal

Developing a User-Centered Mobile Application Prototype: Bridging Lower-Limb Fracture Care from Skilled Nursing Facility and Back to the Community.

AMIA ... Annual Symposium proceedings. AMIA Symposium·2026
Same journal

KERAP: A Knowledge-Enhanced Reasoning Approach for Accurate Zero-shot Diagnosis Prediction Using Multi-agent LLMs.

AMIA ... Annual Symposium proceedings. AMIA Symposium·2026
Same journal

Automating Adjudication of Cardiovascular Events Using Large Language Models.

AMIA ... Annual Symposium proceedings. AMIA Symposium·2026
Same journal

Predictive Factors and State-Level Barriers to Postpartum Birth Control Usage in the United States: Insights from PRAMS Phase 8.

AMIA ... Annual Symposium proceedings. AMIA Symposium·2026
See all related articles

Deep learning models offer powerful healthcare predictions but lack interpretability. Our interpretable mimic learning method uses gradient boosting trees to provide accurate and understandable clinical insights from electronic health records.

Area of Science:

  • Biomedical Informatics
  • Machine Learning in Healthcare
  • Clinical Data Science

Background:

  • The exponential growth of healthcare data, including electronic health records (EHR) and intensive care unit (ICU) sensor data, presents opportunities for data-driven disease pattern discovery.
  • Deep learning models achieve state-of-the-art performance in computational phenotyping and healthcare prediction but suffer from a lack of interpretability, hindering their clinical adoption.
  • Interpretability is crucial for medical research and clinical decision-making, necessitating methods that balance predictive power with transparency.

Purpose of the Study:

  • To introduce interpretable mimic learning, a novel knowledge distillation approach.
  • To develop a method that leverages gradient boosting trees for interpretable modeling while maintaining high prediction accuracy.
  • To address the interpretability gap in deep learning for healthcare applications.

Related Experiment Videos

Main Methods:

  • Employing a knowledge distillation framework to transfer insights from complex models to simpler, interpretable ones.
  • Utilizing gradient boosting trees as the base for learning interpretable models.
  • Validating the approach on a Pediatric ICU dataset for acute lung injury (ALI) prediction tasks.

Main Results:

  • The proposed interpretable mimic learning method achieves strong prediction performance comparable to deep learning models.
  • Experimental results demonstrate superior performance over state-of-the-art approaches for predicting mortality and ventilator-free days.
  • The method successfully generates interpretable models, offering valuable insights to clinicians.

Conclusions:

  • Interpretable mimic learning offers a viable solution to the interpretability challenge in deep learning for healthcare.
  • The approach enhances clinical decision-making by providing both accurate predictions and understandable model explanations.
  • This method holds significant potential for advancing data-driven healthcare research and practice.