Hugging Face's logo

monai-test
/
swin_unetr_btcv_segmentation

monai
medical
Model card Files Community
katielink commited on
Commit
93534f3
•
1 Parent(s): 1cd32b1

Update figure and benchmarking

Browse files
Files changed (3) hide show
  1. README.md +14 -3
  2. configs/metadata.json +2 -1
  3. docs/README.md +14 -3
README.md CHANGED
@@ -10,6 +10,11 @@ A pre-trained model for volumetric (3D) multi-organ segmentation from CT image.
10
 
11
  # Model Overview
12
  A pre-trained Swin UNETR [1,2] for volumetric (3D) multi-organ segmentation using CT images from Beyond the Cranial Vault (BTCV) Segmentation Challenge dataset [3].
 
 
 
 
 
13
  ## Data
14
  The training data is from the [BTCV dataset](https://www.synapse.org/#!Synapse:syn3193805/wiki/89480/) (Please regist in `Synapse` and download the `Abdomen/RawData.zip`).
15
  The dataset format needs to be redefined using the following commands:
@@ -34,16 +39,22 @@ Actual Model Input: 96 x 96 x 96
34
  ## Input and output formats
35
  Input: 1 channel CT image
36
 
37
- Output: 14 channels: 0:Background, 1:Spleen, 2:Right Kidney, 3:Left Kideny, 4:Gallbladder, 5:Esophagus, 6:Liver, 7:Stomach, 8:Aorta, 9:IVC, 10:Portal and Splenic Veins, 11:Pancreas, 12:Right adrenal gland, 13:Left adrenal gland
38
 
39
  ## Performance
 
 
 
 
 
 
40
  A graph showing the validation mean Dice for 5000 epochs.
41
 
42
- ![](https://developer.download.nvidia.com/assets/Clara/Images/monai_swin_unetr_segmentation_val.png) <br>
43
 
44
  This model achieves the following Dice score on the validation data (our own split from the training dataset):
45
 
46
- Mean Dice = 0.8283
47
 
48
  Note that mean dice is computed in the original spacing of the input data.
49
  ## commands example
 
10
 
11
  # Model Overview
12
  A pre-trained Swin UNETR [1,2] for volumetric (3D) multi-organ segmentation using CT images from Beyond the Cranial Vault (BTCV) Segmentation Challenge dataset [3].
13
+
14
+ The architecture of Swin UNETR is shown as below:
15
+
16
+ ![image](https://developer.download.nvidia.com/assets/Clara/Images/monai_swin_unetr_btcv_segmentation_workflow_v1.png)
17
+
18
  ## Data
19
  The training data is from the [BTCV dataset](https://www.synapse.org/#!Synapse:syn3193805/wiki/89480/) (Please regist in `Synapse` and download the `Abdomen/RawData.zip`).
20
  The dataset format needs to be redefined using the following commands:
 
39
  ## Input and output formats
40
  Input: 1 channel CT image
41
 
42
+ Output: 14 channels: 0:Background, 1:Spleen, 2:Right Kidney, 3:Left Kidney, 4:Gallbladder, 5:Esophagus, 6:Liver, 7:Stomach, 8:Aorta, 9:IVC, 10:Portal and Splenic Veins, 11:Pancreas, 12:Right adrenal gland, 13:Left adrenal gland
43
 
44
  ## Performance
45
+
46
+ The figure shows the training loss curve for 10K iterations.
47
+
48
+
49
+ <p align = "center"><img src="https://developer.download.nvidia.com/assets/Clara/Images/monai_swin_unetr_btcv_segmentation_trainloss_v1.png" alt="drawing" width="700"/></p>
50
+
51
  A graph showing the validation mean Dice for 5000 epochs.
52
 
53
+ <p align = "center"><img src="https://developer.download.nvidia.com/assets/Clara/Images/monai_swin_unetr_btcv_segmentation_validation_meandice_v1.png" alt="drawing" width="700"/></p>
54
 
55
  This model achieves the following Dice score on the validation data (our own split from the training dataset):
56
 
57
+ Mean Dice = 0.8269
58
 
59
  Note that mean dice is computed in the original spacing of the input data.
60
  ## commands example
configs/metadata.json CHANGED
@@ -1,7 +1,8 @@
1
  {
2
  "schema": "https://github.com/Project-MONAI/MONAI-extra-test-data/releases/download/0.8.1/meta_schema_20220324.json",
3
- "version": "0.3.4",
4
  "changelog": {
 
5
  "0.3.4": "Update figure link in readme",
6
  "0.3.3": "Update, verify MONAI 1.0.1 and Pytorch 1.13.0",
7
  "0.3.2": "enhance readme on commands example",
 
1
  {
2
  "schema": "https://github.com/Project-MONAI/MONAI-extra-test-data/releases/download/0.8.1/meta_schema_20220324.json",
3
+ "version": "0.3.5",
4
  "changelog": {
5
+ "0.3.5": "Update figure and benchmarking",
6
  "0.3.4": "Update figure link in readme",
7
  "0.3.3": "Update, verify MONAI 1.0.1 and Pytorch 1.13.0",
8
  "0.3.2": "enhance readme on commands example",
docs/README.md CHANGED
@@ -3,6 +3,11 @@ A pre-trained model for volumetric (3D) multi-organ segmentation from CT image.
3
 
4
  # Model Overview
5
  A pre-trained Swin UNETR [1,2] for volumetric (3D) multi-organ segmentation using CT images from Beyond the Cranial Vault (BTCV) Segmentation Challenge dataset [3].
 
 
 
 
 
6
  ## Data
7
  The training data is from the [BTCV dataset](https://www.synapse.org/#!Synapse:syn3193805/wiki/89480/) (Please regist in `Synapse` and download the `Abdomen/RawData.zip`).
8
  The dataset format needs to be redefined using the following commands:
@@ -27,16 +32,22 @@ Actual Model Input: 96 x 96 x 96
27
  ## Input and output formats
28
  Input: 1 channel CT image
29
 
30
- Output: 14 channels: 0:Background, 1:Spleen, 2:Right Kidney, 3:Left Kideny, 4:Gallbladder, 5:Esophagus, 6:Liver, 7:Stomach, 8:Aorta, 9:IVC, 10:Portal and Splenic Veins, 11:Pancreas, 12:Right adrenal gland, 13:Left adrenal gland
31
 
32
  ## Performance
 
 
 
 
 
 
33
  A graph showing the validation mean Dice for 5000 epochs.
34
 
35
- ![](https://developer.download.nvidia.com/assets/Clara/Images/monai_swin_unetr_segmentation_val.png) <br>
36
 
37
  This model achieves the following Dice score on the validation data (our own split from the training dataset):
38
 
39
- Mean Dice = 0.8283
40
 
41
  Note that mean dice is computed in the original spacing of the input data.
42
  ## commands example
 
3
 
4
  # Model Overview
5
  A pre-trained Swin UNETR [1,2] for volumetric (3D) multi-organ segmentation using CT images from Beyond the Cranial Vault (BTCV) Segmentation Challenge dataset [3].
6
+
7
+ The architecture of Swin UNETR is shown as below:
8
+
9
+ ![image](https://developer.download.nvidia.com/assets/Clara/Images/monai_swin_unetr_btcv_segmentation_workflow_v1.png)
10
+
11
  ## Data
12
  The training data is from the [BTCV dataset](https://www.synapse.org/#!Synapse:syn3193805/wiki/89480/) (Please regist in `Synapse` and download the `Abdomen/RawData.zip`).
13
  The dataset format needs to be redefined using the following commands:
 
32
  ## Input and output formats
33
  Input: 1 channel CT image
34
 
35
+ Output: 14 channels: 0:Background, 1:Spleen, 2:Right Kidney, 3:Left Kidney, 4:Gallbladder, 5:Esophagus, 6:Liver, 7:Stomach, 8:Aorta, 9:IVC, 10:Portal and Splenic Veins, 11:Pancreas, 12:Right adrenal gland, 13:Left adrenal gland
36
 
37
  ## Performance
38
+
39
+ The figure shows the training loss curve for 10K iterations.
40
+
41
+
42
+ <p align = "center"><img src="https://developer.download.nvidia.com/assets/Clara/Images/monai_swin_unetr_btcv_segmentation_trainloss_v1.png" alt="drawing" width="700"/></p>
43
+
44
  A graph showing the validation mean Dice for 5000 epochs.
45
 
46
+ <p align = "center"><img src="https://developer.download.nvidia.com/assets/Clara/Images/monai_swin_unetr_btcv_segmentation_validation_meandice_v1.png" alt="drawing" width="700"/></p>
47
 
48
  This model achieves the following Dice score on the validation data (our own split from the training dataset):
49
 
50
+ Mean Dice = 0.8269
51
 
52
  Note that mean dice is computed in the original spacing of the input data.
53
  ## commands example