Skip to main content
 

Automatic Screening for Lung Diseases in Chest Radiographs: A Global Health Initiative

Thursday, October 11, 2012 — Poster Session III

10:00 a.m. – Noon

Natcher Conference Center, Building 45

NLM

IMAG-12

Authors

  • S. Jaeger
  • A. Karargyris
  • S. Antani
  • G. Thoma

Abstract

We are developing algorithms and software to automatically screen chest x-rays (CXR) for lung diseases, with a special focus on Tuberculosis. TB infects about one-third of the world’s population, with nine million new cases appearing every year, and furthermore is an opportunistic infection affecting immune-compromised HIV/AIDS patients, often in resource-poor regions with limited radiological services. Following development and evaluation, we intend to deploy this software in conjunction with portable x-ray scanners, to screen rural populations in Kenya and to refer infected patients for further attention by a large AIDS treatment program. Our processing stages include: enhancing the contrast of CXR by histogram equalization, extracting the lung areas by an adaptive segmentation method, removing confounding structures such as heart and ribs, detecting abnormalities in the lungs using texture and shape features, and employing a binary support vector machine-based classifier to differentiate normal from abnormal findings. In this paper we describe the image processing and classification steps, system performance results, and the status of deployment in the field.

back to top