Using Hierarchy label classification to Chest X-ray images

Principal Investigator

Name
Gregory Hager

Degrees
Ph.D.

Institution
Johns Hopkins University

Position Title
Professor

Email
hager@cs.jhu.edu

About this CDAS Project

Study
PLCO (Learn more about this study)

Project ID
PLCO-384

Initial CDAS Request Approval
Jul 20, 2018

Title
Using Hierarchy label classification to Chest X-ray images

Summary
Because chest diseases often has relationship with each other, I want to compare the result of hierarchy classification with flat classification on chest x-ray to see whether a relationship-cared tree label can get better results.

Aims

I aim to use both PLCO dataset and ChestX-ray dataset created by Xiaosong to get better results.

Because I used tree-based label, it will learn patterns instead of specific labels more. So it can reduce the influence of variance of radiologists.

I want to get better performance on Chest X-ray classification.

Collaborators

Haomin Chen, Johns Hopkins University

Related Publications

• Deep hiearchical multi-label classification applied to chest X-ray abnormality taxonomies.
  Chen H, Miao S, Xu D, Hager GD, Harrison AP
  Med Image Anal. 2020 Sep 5; Volume 66: Pages 101811 PUBMED