Comparison of Deep Learning and Radiomics CT Image Signatures

Published: 3 September 2020| Version 1 | DOI: 10.17632/yn7vpd7bxk.1
Jiangdian Song


This is a pneumonia CT dataset for the comparison of image signatures constructed by deep learning architecture and radiomics. The region of interest (ROI) of pneumonia on CT images was manually delineated by our radiologists using the ITK-Snap slice-by-slice. The pixel intensity of the pneumonia lesions contained within the ROI was the original CT intensity and the others was set to zero. The images in this HDF5 file was classified into COVID-19 positive and COVID-19 negative. The shape of each slice of CT images is (256, 256), which down-sampling from (512, 512). Please cite the reference if using this dataset: Hongmei Wang, Lu Wang, Edward H. Lee, Jimmy Zheng, Wei Zhang, Safwan Halabi, Chunlei Liu, Kexue Deng, Jiangdian Song, and Kristen W. Yeom. Decoding COVID-19 Pneumonia: Comparison of Deep Learning and Radiomics CT Image Signatures.


Steps to reproduce

The source code has been provided:


Computed Tomography