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

Incremental domain adaptation-based ICU patient mortality prediction.

Xinqiang Xie1, Ting Lyu2,3, Tao Wei1

  • 1College of Artificial Intelligence, Dalian Maritime University, Dalian, Liaoning, China.

Digital Health
|July 9, 2026
PubMed
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

[Adsorption characteristics of acetone and butanone onto honeycomb ZSM-5 molecular sieve].

Huan jing ke xue= Huanjing kexue·2014
Same author

First isolation of reticuloendotheliosis virus from mallards in China.

Archives of virology·2014
Same author

Association analysis of STK39, MCCC1/LAMP3 and sporadic PD in the Chinese Han population.

Neuroscience letters·2014
Same author

Tumor-responsive fluorescent light-up probe based on a gold nanoparticle/conjugated polyelectrolyte hybrid.

Small (Weinheim an der Bergstrasse, Germany)·2014
Same author

Transcriptomic analysis of differentially expressed genes in the Ras1(CA)-overexpressed and wildtype posterior silk glands.

BMC genomics·2014
Same author

Reversible photoswitching conjugated polymer nanoparticles for cell and ex vivo tumor imaging.

Nanoscale·2014

AdaICU enhances intensive care unit (ICU) mortality prediction by adapting models to new data, improving accuracy and reliability in diverse clinical settings. This approach addresses patient heterogeneity and data dynamics for better patient outcomes.

Area of Science:

  • Medical Informatics
  • Machine Learning in Healthcare
  • Clinical Prediction Models

Background:

  • Accurate ICU mortality prediction is vital for timely interventions.
  • Existing models struggle with patient heterogeneity and dynamic ICU data, limiting generalization.
  • Dynamically adaptive predictive methods are crucial for real-world clinical reliability.

Purpose of the Study:

  • To develop a novel framework, AdaICU, for cross-ICU mortality prediction.
  • To integrate unsupervised domain adaptation and incremental learning for improved model adaptability.
  • To address limitations in generalization and self-updating capabilities of current prediction models.

Main Methods:

  • AdaICU utilizes unsupervised domain adaptation to align feature distributions between source and target domains.
Keywords:
ICUdeep learningdomain adaptationmortality prediction

Related Experiment Videos

  • Domain-adversarial training is employed for feature distribution alignment.
  • Incremental learning is used to update the model with new target-domain data.

    • Main Results:

    • AdaICU consistently outperformed baseline models in cross-domain settings on the PhysioNet 2012 dataset.
    • Achieved 2-6% AUROC improvement and substantial AUPRC gains over baselines.
    • Demonstrated 4-8% accuracy improvement and higher positive predictive value.

    Conclusions:

    • AdaICU effectively addresses cross-domain distribution shifts and enables post-deployment adaptation.
    • Provides a practical framework for ICU mortality prediction with limited labeled data.
    • Shows promise for dynamic clinical environments with evolving data distributions.