Skip to content

Rajaraman et al. ResNet-BS chest radiograph bone suppression deep learning network

Notifications You must be signed in to change notification settings

danielnflam/CXR-bone-suppression

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

65 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Pytorch implementation of Rajaraman's work

Rajaraman, S.; Zamzmi, G.; Folio, L.; Alderson, P.; Antani, S. Chest X-Ray Bone Suppression for Improving Classification of Tuberculosis-Consistent Findings. Diagnostics 2021, 11, 840. https://doi.org/10.3390/diagnostics11050840

A pre-trained model with the weights of the ResNet-BS bone suppression model is included for direct use. The pre-trained network is trained on the JSRT and BSE-JSRT dataset, consisting of 177 training image pairs, 20 validation pairs and 20 internal test pairs. Run the model using analysis_script.ipynb on 256 x 256 grayscale CXR image to generate a soft-tissue image with suppressed bone shadows.

pytorch-msssim is sourced from: https://github.com/jorge-pessoa/pytorch-msssim

Training:

Use the main.ipynb script. The settings to change are in the first 2 cells.

Inference:

Use the analysis_script.ipynb

Your own datasets:

Put your own datasets in datasets.py. Pre-existing dataset classes are provided as templates.

About

Rajaraman et al. ResNet-BS chest radiograph bone suppression deep learning network

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published