CT Chest Segmentation Dataset.
This dataset was be modified from Lung segmentation dataset by Kónya et al., 2020 , https://www.kaggle.com/sandorkonya/ct-lung-heart-trachea-segmentation
The original nrrd files were re-saved in single tensor format with masks corresponding to labels: (lungs, heart, trachea) as numpy arrays using pickle.
Each tensor has the following shape: number of slices, width, height, number of classes, where the width and height number of slices are individual parameters of each tensor id, and number of classes = 3.
In addition, the data was re-saved as RGB images, where each image corresponds to one ID slice, and their mask-images have channels corresponding to three classes: (lung, heart, trachea).
The dataset contains:
- numpy_images_files.zip - images in numpy format.
- numpy_masks_files.zip - segmentation masks in numpy format.
- images.zip - images in RGB format.
- masks.zip - segmentation masks in RGB format.
- train.csv - csv file with image names.
Below is an example of what the data looks like:
- .npy files can be readed like this:
import pickle
with open(file_path, 'rb') as f:
tensor = pickle.load(f)
- The images look like this:
The code with dataset creation available here - dataset_creation.ipynb