|
a |
|
b/dataset_description.md |
|
|
1 |
### Dataset Description |
|
|
2 |
|
|
|
3 |
CT Chest Segmentation Dataset. |
|
|
4 |
This dataset was be modified from `Lung segmentation dataset by Kónya et al., 2020 , https://www.kaggle.com/sandorkonya/ct-lung-heart-trachea-segmentation` |
|
|
5 |
|
|
|
6 |
The original nrrd files were re-saved in single tensor format with masks corresponding to labels: (**lungs**, **heart**, **trachea**) as numpy arrays using pickle. |
|
|
7 |
|
|
|
8 |
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. |
|
|
9 |
|
|
|
10 |
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**). |
|
|
11 |
|
|
|
12 |
<p> |
|
|
13 |
<img src="https://github.com/mandrakedrink/ChestCTSegmentation/blob/master/stats/data/demo.gif" width="35%" height="35%"> |
|
|
14 |
</p> |
|
|
15 |
|
|
|
16 |
|
|
|
17 |
|
|
|
18 |
### Content |
|
|
19 |
**The dataset contains**: |
|
|
20 |
|
|
|
21 |
+ numpy_images_files.zip - images in numpy format. |
|
|
22 |
+ numpy_masks_files.zip - segmentation masks in numpy format. |
|
|
23 |
+ images.zip - images in RGB format. |
|
|
24 |
+ masks.zip - segmentation masks in RGB format. |
|
|
25 |
+ train.csv - csv file with image names. |
|
|
26 |
|
|
|
27 |
|
|
|
28 |
**Below is an example of what the data looks like:** |
|
|
29 |
|
|
|
30 |
+ **.npy files can be readed like this:** |
|
|
31 |
|
|
|
32 |
``` |
|
|
33 |
import pickle |
|
|
34 |
|
|
|
35 |
with open(file_path, 'rb') as f: |
|
|
36 |
tensor = pickle.load(f) |
|
|
37 |
``` |
|
|
38 |
|
|
|
39 |
+ **The images look like this:** |
|
|
40 |
<p> |
|
|
41 |
<img src="https://github.com/mandrakedrink/ChestCTSegmentation/blob/master/stats/data/svg/data_example2.svg" width="60%" height="60%"> |
|
|
42 |
</p> |
|
|
43 |
|
|
|
44 |
<p> |
|
|
45 |
<img src="https://github.com/mandrakedrink/ChestCTSegmentation/blob/master/stats/data/svg/data_example1.svg" width="60%" height="60%"> |
|
|
46 |
</p> |
|
|
47 |
|
|
|
48 |
|
|
|
49 |
The code with dataset creation available here - [dataset_creation.ipynb](https://github.com/mandrakedrink/ChestCTSegmentation/blob/master/dataset_creation.ipynb)[<img src="https://colab.research.google.com/assets/colab-badge.svg" align="center">](https://colab.research.google.com/drive/166TOgOsRvcblQK2j_HTB8CmVy5VGabas?usp=sharing) |
|
|
50 |
|
|
|
51 |
|
|
|
52 |
--- |