Images of COVID-19 positive and negative pneumonia patients

Published: 3 September 2020| Version 2 | DOI: 10.17632/kk6y7nnbfs.2
Contributor:
Jiangdian Song

Description

We uploaded the CT slices with suspected COVID-19 pneumonia signs marked by radiologists for each patient (both positive and negative) in this study to ensure that all patients' images can be uploaded within this capacity. Among them, CT images (training) from one hospital that can be divided into the training, validation, and test datasets are stored in the "Data_For_Training_Validation_Test.hdf5" file. The shape of each slice of CT images is (256, 256), down-sampling from (512, 512) in order to save space. The CT images from the other hospital used for external validation dataset is stored in the "ExternalValidationData.hdf5" file. The shape of each layer of CT images is (256, 256, 3). The data dimension should be extended to three when using BigBiGAN framework to extract image semantic features. In this dataset, in order to save space, we did not expand the training dataset. BigBiGAN training could be implemented using the source code we provide: https://github.com/MI-12/BigBIGAN-for-COVID-19. with h5py.File("TrainData.hdf5", 'r') as record: keys = list(record.keys()) The above two lines of code are used to obtain the keys of each image, that is, the name of each image. The name of the CT image is corresponds to the name of the image in the excel files provided. Please cite the reference if using this dataset: Jiangdian Song, Hongmei Wang, Yuchan Liu, Wenqing Wu, Gang Dai, Zongshan Wu, Puhe Zhu, Wei Zhang, Keisten W. Yeom, Kexue Deng. End-to-end automatic differentiation of the coronavirus disease 2019 (COVID-19) from viral pneumonia based on chest CT. European journal of nuclear medicine and molecular imaging (2020): 1-9.

Files

Steps to reproduce

The code has been published at: https://github.com/MI-12/BigBIGAN-for-COVID-19.