GoodBaiBai88
commited on
Update README.md
Browse files
README.md
CHANGED
@@ -1,33 +1,53 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
tags:
|
4 |
-
-
|
5 |
- 3D medical segmentation
|
6 |
size_categories:
|
7 |
- 1K<n<10K
|
8 |
---
|
9 |
|
|
|
|
|
10 |
## Dataset Description
|
11 |
Large-scale General 3D Medical Image Segmentation Dataset (M3D-Seg)
|
12 |
|
13 |
### Dataset Introduction
|
14 |
-
3D medical segmentation
|
15 |
-
Currently, due to privacy and cost
|
16 |
-
To address this, we have collected 25 publicly available
|
17 |
-
including CHAOS, HaN-Seg, AMOS22, AbdomenCT-1k, KiTS23, KiPA22,
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
|
27 |
### Supported Tasks
|
28 |
-
|
29 |
-
|
30 |
-
|
|
|
|
|
|
|
|
|
31 |
|
32 |
## Dataset Format and Structure
|
33 |
|
@@ -35,48 +55,71 @@ the dataset supports tasks such as: 3D segmentation: semantic segmentation, text
|
|
35 |
<pre>
|
36 |
M3D_Seg/
|
37 |
0000/
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
......
|
44 |
0000.json
|
45 |
0001/
|
46 |
......
|
47 |
</pre>
|
48 |
|
49 |
### Dataset Download
|
|
|
|
|
50 |
#### Clone with HTTP
|
51 |
```bash
|
52 |
-
git clone
|
|
|
|
|
|
|
|
|
|
|
|
|
53 |
```
|
54 |
#### Manual Download
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
60 |
|
61 |
|
62 |
### Dataset Loading Method
|
63 |
-
#### 1.
|
64 |
-
|
65 |
-
|
66 |
-
Please
|
|
|
|
|
|
|
|
|
67 |
|
68 |
#### 2. Build Dataset
|
69 |
-
|
|
|
70 |
|
71 |
```python
|
72 |
|
73 |
```
|
74 |
|
75 |
|
|
|
|
|
|
|
76 |
|
77 |
|
78 |
-
### Data Splitting
|
79 |
-
Each file is split into ‘train, validation/test’ using json files, for ease of training and testing models.
|
80 |
|
81 |
### Dataset Sources
|
82 |
|
@@ -89,7 +132,7 @@ Each file is split into ‘train, validation/test’ using json files, for ease
|
|
89 |
| 0004 |KiTS23| https://kits-challenge.org/kits23/|
|
90 |
| 0005 |KiPA22| https://kipa22.grand-challenge.org/|
|
91 |
| 0006 |KiTS19| https://kits19.grand-challenge.org/|
|
92 |
-
| 0007 |BTCV| https://www.synapse.org
|
93 |
| 0008 |Pancreas-CT| https://wiki.cancerimagingarchive.net/display/public/pancreas-ct|
|
94 |
| 0009 | 3D-IRCADB | https://www.kaggle.com/datasets/nguyenhoainam27/3dircadb |
|
95 |
| 0010 |FLARE22| https://flare22.grand-challenge.org/|
|
@@ -111,7 +154,7 @@ Each file is split into ‘train, validation/test’ using json files, for ease
|
|
111 |
|
112 |
## Dataset Copyright Information
|
113 |
|
114 |
-
All datasets
|
115 |
|
116 |
## Citation
|
117 |
If you use this dataset, please cite the following works:
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
tags:
|
4 |
+
- multi-modal
|
5 |
- 3D medical segmentation
|
6 |
size_categories:
|
7 |
- 1K<n<10K
|
8 |
---
|
9 |
|
10 |
+
![Data_visualization](M3D_Seg.jpg)
|
11 |
+
|
12 |
## Dataset Description
|
13 |
Large-scale General 3D Medical Image Segmentation Dataset (M3D-Seg)
|
14 |
|
15 |
### Dataset Introduction
|
16 |
+
3D medical image segmentation poses a significant challenge in medical image analysis.
|
17 |
+
Currently, due to privacy and cost constraints, publicly available large-scale 3D medical images
|
18 |
+
and their annotated data are scarce. To address this, we have collected 25 publicly available
|
19 |
+
3D CT segmentation datasets, including CHAOS, HaN-Seg, AMOS22, AbdomenCT-1k, KiTS23, KiPA22,
|
20 |
+
KiTS19, BTCV, Pancreas-CT, 3D-IRCADB, FLARE22, TotalSegmentator, CT-ORG, WORD, VerSe19, VerSe20,
|
21 |
+
SLIVER07, QUBIQ, MSD-Colon, MSD-HepaticVessel, MSD-Liver, MSD-lung, MSD-pancreas, MSD-spleen, LUNA16.
|
22 |
+
These datasets are uniformly encoded from 0000 to 0024, totaling 5,772 3D images and 149,196 3D mask annotations.
|
23 |
+
The semantic labels corresponding to each mask can be represented in text.
|
24 |
+
Within each sub-dataset folder, there are multiple data folders (containing image and mask files),
|
25 |
+
and each sub-dataset independently utilizes its JSON file to split.
|
26 |
+
|
27 |
+
- **data_load_demo.py**: Provides an example code on reading images and masks from the dataset.
|
28 |
+
- **data_preocess.py**: Describes how to convert raw `nii.gz` or other format data into a more efficient `npy` format and preprocess them,
|
29 |
+
saving them in a unified format. This dataset has already been preprocessed, so there is no need to use data_preocess.py again.
|
30 |
+
If adding new datasets, please follow a unified processing approach.
|
31 |
+
- **dataset_info.json & dataset_info.txt**: Contain the names of each dataset and their label texts.
|
32 |
+
- **term_dictionary.json**: Provides multiple definitions or descriptions for each semantic label in the dataset,
|
33 |
+
generated by `ChatGPT` for each term. Researchers can convert category IDs in the dataset to label texts
|
34 |
+
using the information in dataset_info.txt and further convert them into text descriptions using term_dictionary.json as text inputs for segmentation models,
|
35 |
+
enabling tasks such as segmentation based on text prompts and referring segmentation.
|
36 |
+
|
37 |
+
This dataset supports not only traditional semantic segmentation tasks but also text-based segmentation tasks.
|
38 |
+
For detailed methods, please refer to [SegVol](https://github.com/BAAI-DCAI/SegVol) and [M3D](https://github.com/BAAI-DCAI/M3D).
|
39 |
+
As a general segmentation dataset, we provide a convenient, unified, and structured dataset organization
|
40 |
+
that allows for the uniform integration of more public and private datasets in the same format as this dataset,
|
41 |
+
thereby constructing a larger-scale general 3D medical image segmentation dataset.
|
42 |
|
43 |
### Supported Tasks
|
44 |
+
This dataset not only supports traditional image-mask semantic segmentation tasks
|
45 |
+
but also represents data in the form of image-mask-text, where masks can be converted into box coordinates
|
46 |
+
through bounding boxes. Based on this, the dataset can effectively support a series of image segmentation
|
47 |
+
and positioning tasks, as follows:
|
48 |
+
|
49 |
+
- **3D Segmentation**: Semantic segmentation, text-based segmentation, referring segmentation, reasoning segmentation, etc.
|
50 |
+
- **3D Positioning**: Visual grounding/referring expression comprehension, referring expression generation.
|
51 |
|
52 |
## Dataset Format and Structure
|
53 |
|
|
|
55 |
<pre>
|
56 |
M3D_Seg/
|
57 |
0000/
|
58 |
+
1/
|
59 |
+
image.npy
|
60 |
+
mask_(1, 512, 512, 96).npz
|
61 |
+
2/
|
62 |
+
......
|
|
|
63 |
0000.json
|
64 |
0001/
|
65 |
......
|
66 |
</pre>
|
67 |
|
68 |
### Dataset Download
|
69 |
+
The total dataset size is approximately 224G.
|
70 |
+
|
71 |
#### Clone with HTTP
|
72 |
```bash
|
73 |
+
git clone https://huggingface.co/datasets/GoodBaiBai88/M3D-Seg
|
74 |
+
```
|
75 |
+
|
76 |
+
#### SDK Download
|
77 |
+
```bash
|
78 |
+
from datasets import load_dataset
|
79 |
+
dataset = load_dataset("GoodBaiBai88/M3D-Seg")
|
80 |
```
|
81 |
#### Manual Download
|
82 |
+
Manually download all files from the dataset files. It is recommended to use batch download tools for efficient downloading.
|
83 |
+
Please note the following:
|
84 |
+
|
85 |
+
- **Downloading in Parts and Merging**: Since dataset 0024 has a large volume,
|
86 |
+
the original compressed file has been split into two parts: `0024_1` and `0024_2`.
|
87 |
+
Make sure to download these two files separately and unzip them in the same directory to ensure data integrity.
|
88 |
+
|
89 |
+
- **Masks with Sparse Matrices**: To save storage space effectively,
|
90 |
+
foreground information in masks is stored in sparse matrix format and saved with the extension `.npz`.
|
91 |
+
The name of each mask file typically includes its shape information for identification and loading purposes.
|
92 |
+
|
93 |
+
- **Data Load Demo**: There is a script named data_load_demo.py, which serves as a reference for correctly
|
94 |
+
reading the sparse matrix format of masks and other related data.
|
95 |
+
Please refer to this script for specific loading procedures and required dependencies.
|
96 |
+
|
97 |
|
98 |
|
99 |
### Dataset Loading Method
|
100 |
+
#### 1. Direct Usage of Preprocessed Data
|
101 |
+
If you have already downloaded the preprocessed dataset, no additional data processing steps are required.
|
102 |
+
You can directly jump to step 2 to build and load the dataset.
|
103 |
+
Please note that the contents provided by this dataset have been transformed and numbered through data_process.py,
|
104 |
+
differing from the original `nii.gz` files. To understand the specific preprocessing process,
|
105 |
+
refer to the data_process.py file for detailed information. If adding new datasets or modifying existing ones,
|
106 |
+
please refer to data_process.py for data preprocessing and uniform formatting.
|
107 |
+
|
108 |
|
109 |
#### 2. Build Dataset
|
110 |
+
To facilitate model training and evaluation using this dataset, we provide an example code for the Dataset class.
|
111 |
+
Wrap the dataset in your project according to the following example:
|
112 |
|
113 |
```python
|
114 |
|
115 |
```
|
116 |
|
117 |
|
118 |
+
### Data Splitting
|
119 |
+
Each sub-dataset folder is splitted into `train` and `test` parts through a JSON file,
|
120 |
+
facilitating model training and testing.
|
121 |
|
122 |
|
|
|
|
|
123 |
|
124 |
### Dataset Sources
|
125 |
|
|
|
132 |
| 0004 |KiTS23| https://kits-challenge.org/kits23/|
|
133 |
| 0005 |KiPA22| https://kipa22.grand-challenge.org/|
|
134 |
| 0006 |KiTS19| https://kits19.grand-challenge.org/|
|
135 |
+
| 0007 |BTCV| https://www.synapse.org/#!Synapse:syn3193805/wiki/217753|
|
136 |
| 0008 |Pancreas-CT| https://wiki.cancerimagingarchive.net/display/public/pancreas-ct|
|
137 |
| 0009 | 3D-IRCADB | https://www.kaggle.com/datasets/nguyenhoainam27/3dircadb |
|
138 |
| 0010 |FLARE22| https://flare22.grand-challenge.org/|
|
|
|
154 |
|
155 |
## Dataset Copyright Information
|
156 |
|
157 |
+
All datasets in this dataset are publicly available. For detailed copyright information, please refer to the corresponding dataset links.
|
158 |
|
159 |
## Citation
|
160 |
If you use this dataset, please cite the following works:
|