Bone Structures Extraction and Enhancement in Chest Radiographs via CNN Trained on Synthetic Data

Ophir Gozes, Hayit Greenspan

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

4 Scopus citations

Abstract

In this paper, we present a deep learning-based image processing technique for extraction of bone structures in chest radiographs using a U-Net FCNN. The U-Net was trained to accomplish the task in a fully supervised setting. To create the training image pairs, we employed simulated X-Ray or Digitally Reconstructed Radiographs (DRR), derived from 664 CT scans belonging to the LIDC-IDRI dataset. Using HU based segmentation of bone structures in the CT domain, a synthetic 2D 'Bone x-ray' DRR is produced and used for training the network. For the reconstruction loss, we utilize two loss functions- L1 Loss and perceptual loss. Once the bone structures are extracted, the original image can be enhanced by fusing the original input x-ray and the synthesized 'Bone X-ray'. We show that our enhancement technique is applicable to real x-ray data, and display our results on the NIH Chest X-Ray-14 dataset.

Original languageEnglish
Title of host publicationISBI 2020 - 2020 IEEE International Symposium on Biomedical Imaging
PublisherIEEE Computer Society
Pages858-861
Number of pages4
ISBN (Electronic)9781538693308
DOIs
StatePublished - Apr 2020
Externally publishedYes
Event17th IEEE International Symposium on Biomedical Imaging, ISBI 2020 - Iowa City, United States
Duration: 3 Apr 20207 Apr 2020

Publication series

NameProceedings - International Symposium on Biomedical Imaging
Volume2020-April
ISSN (Print)1945-7928
ISSN (Electronic)1945-8452

Conference

Conference17th IEEE International Symposium on Biomedical Imaging, ISBI 2020
Country/TerritoryUnited States
CityIowa City
Period3/04/207/04/20

Keywords

  • CT
  • DRR
  • Deep learning
  • Image enhancement
  • Image synthesis
  • X-ray

Fingerprint

Dive into the research topics of 'Bone Structures Extraction and Enhancement in Chest Radiographs via CNN Trained on Synthetic Data'. Together they form a unique fingerprint.

Cite this