This repository houses a sophisticated Conditional Generative Adversarial Network (cGAN) designed to generate synthetic chest X-ray images, depicting six distinct medical conditions. The primary aim of this endeavor is to expand existing datasets with high-fidelity synthetic images, thereby enhancing the robustness of machine learning models used in medical diagnostics.
The NIH Chest X-ray Dataset serves as the backbone for this project, comprising 112,120 X-ray images from 30,805 patients. Each image is labeled for various disease classes, derived using Natural Language Processing (NLP) on accompanying radiological reports.
For a thorough understanding of the dataset and labeling process, please refer to:
Wang X, Peng Y, Lu L, Lu Z, Bagheri M, Summers RM. "ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases." in IEEE CVPR, 2017.
- NLP-extracted labels with an estimated accuracy above 90%.
- Bounding box annotations are limited in quantity.
- The radiology reports are not included in the dataset.
The dataset is organized into multiple ZIP archives, each containing a segment of the full image collection, accompanied by a README_ChestXray.pdf
for detailed dataset information, BBox_list_2017.csv
for bounding box coordinates, and Data_entry_2017.csv
for class labels and patient data.
There are 15 classes, including 14 disease types and one 'No findings' category. Diseases include Atelectasis, Consolidation, Infiltration, and more. But here for our research we have used 6 different dieases. In total of 7 classes, one being "no finding"
tensorflow
numpy
matplotlib
google.colab
opencv-python
IPython
scikit-image
ipykernel
Loss Graph
This project has been made possible through the resources provided by the NIH Clinical Center and the National Library of Medicine. We extend our gratitude to the dataset authors for making it publicly available.
Please cite the following work when using this dataset or the cGAN model: @inproceedings{wang2017chestx, title={ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases}, author={Wang, X. and Peng, Y. and Lu, L. and Lu, Z. and Bagheri, M. and Summers, R.M.}, booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)}, year={2017} }