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

Adversarial Time-to-Event Modeling | JoVE Visualize

Home
Research Domains
Information And Computing Sciences

Adversarial Time-to-Event Modeling

Paidamoyo Chapfuwa¹, Chenyang Tao¹, Chunyuan Li¹

¹Duke University.

Proceedings of Machine Learning Research|April 9, 2021

View abstract on PubMed

Summary

This study introduces a novel deep learning method for time-to-event analysis, improving the estimation of event-time distributions using censored data. The approach offers significant performance gains over traditional parametric models in health data science.

Area of Science:

Health Data Science
Machine Learning
Biostatistics

Background:

Modern health data science utilizes molecular and electronic health records for clinical practice.
Time-to-event analysis (survival analysis) is a key statistical modeling technique in healthcare.
Accurate estimation of event-time distributions is crucial for clinical decision-making.

Purpose of the Study:

To develop a deep-network-based approach for nonparametric estimation of event-time distributions.
To address the challenge of utilizing information from censored events in time-to-event modeling.
To propose a novel cost function for improved survival analysis.

Main Methods:

Leveraging adversarial learning within a deep network architecture.
Developing a principled cost function to incorporate censored event data.
Focusing on the estimation of time-to-event distributions, not just event ordering.

Main Results:

The proposed deep learning model significantly outperforms a parametric alternative.
Demonstrated performance gains on both benchmark and real-world health datasets.
Successfully estimated nonparametric event-time distributions, including information from censored events.

Conclusions:

The novel deep network approach with adversarial learning enhances time-to-event analysis.
The proposed cost function effectively utilizes censored data for improved distribution estimation.
This method advances statistical modeling in health data science for clinical support.

Machine Learning

Adversarial Machine Learning

Adversarial Time-to-event Modeling