Image-to-Image
MedVAE

Add pipeline tag and library name

#1
by nielsr HF staff - opened
Files changed (1) hide show
  1. README.md +62 -61
README.md CHANGED
@@ -1,61 +1,62 @@
1
- ---
2
- license: mit
3
- ---
4
-
5
- # MedVAE
6
-
7
- MedVAE is a family of six large-scale, generalizable 2D and 3D variational autoencoders (VAEs) designed for medical imaging. It is trained on over one million medical images across multiple anatomical regions and modalities. MedVAE autoencoders encode medical images as downsized latent representations and decode latent representations back to high-resolution images. Across diverse tasks obtained from 20 medical image datasets, we demonstrate that utilizing MedVAE latent representations in place of high-resolution images when training downstream models can lead to efficiency benefits (up to 70x improvement in throughput) while simultaneously preserving clinically-relevant features.
8
-
9
- [[πŸ’» Github](https://github.com/StanfordMIMI/MedVAE)]
10
-
11
- ## πŸ“Œ Model Description
12
- | Total Compression Factor | Channels | Dimensions | Modalities | Anatomies | Config File | Model File |
13
- |----------|----------|----------|----------|----------|----------|----------|
14
- | 16 | 1 | 2D | X-ray | Chest, Breast (FFDM) | [medvae_4x1.yaml ](model_weights/medvae_4x1.yaml)| [vae_4x_1c_2D.ckpt](model_weights/vae_4x_1c_2D.ckpt)
15
- | 16 | 3 | 2D | X-ray | Chest, Breast (FFDM) | [medvae_4x3.yaml](model_weights/medvae_4x3.yaml) | [vae_4x_3c_2D.ckpt](model_weights/vae_4x_3c_2D.ckpt)
16
- | 64 | 1 | 2D | X-ray | Chest, Breast (FFDM) | [medvae_8x1.yaml](model_weights/medvae_8x1.yaml) | [vae_8x_1c_2D.ckpt](model_weights/vae_8x_1c_2D.ckpt)
17
- | 64 | 3 | 2D | X-ray | Chest, Breast (FFDM) | [medvae_8x4.yaml](model_weights/medvae_8x4.yaml) | [vae_8x_4c_2D.ckpt](model_weights/vae_8x_4c_2D.ckpt)
18
- | 64 | 1 | 3D | MRI, CT | Whole-Body | [medvae_4x1.yaml ](model_weights/medvae_4x1.yaml) | [vae_4x_1c_3D.ckpt](model_weights/vae_4x_1c_3D.ckpt)
19
- | 512 | 1 | 3D | MRI, CT | Whole-Body | [medvae_8x1.yaml](model_weights/medvae_8x1.yaml) | [vae_8x_1c_3D.ckpt](model_weights/vae_8x_1c_3D.ckpt)
20
-
21
- Note: Model weights and checkpoints are located in the `model_weights` folder.
22
-
23
- ## ⚑️ Installation
24
- To install MedVAE, you can simply run:
25
-
26
- ```python
27
- pip install medvae
28
- ```
29
-
30
- For an editable installation, use the following commands to clone and install this repository.
31
- ```python
32
- git clone https://github.com/StanfordMIMI/MedVAE.git
33
- cd medvae
34
- pip install -e .[dev]
35
- ```
36
-
37
-
38
- ## πŸš€ Usage Instruction
39
-
40
- We allow for easy-to-use CLI inference tool for compressing your high-dimensional medical images into usable latents:
41
-
42
- ```python
43
- medvae_inference -i INPUT_FOLDER -o OUTPUT_FOLDER -model_name MED_VAE_MODEL -modality MODALITY
44
- ```
45
-
46
- ## πŸ“Ž Citation
47
- If you use MedVAE, please cite the original paper:
48
-
49
- ```bibtex
50
- @article{varma2025medvae,
51
- title = {MedVAE: Efficient Automated Interpretation of Medical Images with Large-Scale Generalizable Autoencoders},
52
- author = {Maya Varma, Ashwin Kumar, Rogier van der Sluijs, Sophie Ostmeier, Louis Blankemeier, Pierre Chambon, Christian Bluethgen, Jip Prince, Curtis Langlotz, Akshay Chaudhari},
53
- year = {2025},
54
- publisher = {Github},
55
- journal = {Github},
56
- howpublished = {https://github.com/StanfordMIMI/MedVAE}
57
- }
58
- ```
59
-
60
- For questions, please place a Github Issues message.
61
-
 
 
1
+ ---
2
+ license: mit
3
+ library_name: medvae
4
+ pipeline_tag: image-to-image
5
+ ---
6
+
7
+ # MedVAE
8
+
9
+ MedVAE is a family of six large-scale, generalizable 2D and 3D variational autoencoders (VAEs) designed for medical imaging. It is trained on over one million medical images across multiple anatomical regions and modalities. MedVAE autoencoders encode medical images as downsized latent representations and decode latent representations back to high-resolution images. Across diverse tasks obtained from 20 medical image datasets, we demonstrate that utilizing MedVAE latent representations in place of high-resolution images when training downstream models can lead to efficiency benefits (up to 70x improvement in throughput) while simultaneously preserving clinically-relevant features.
10
+
11
+ [[πŸ’» Github](https://github.com/StanfordMIMI/MedVAE)]
12
+
13
+ ## πŸ“Œ Model Description
14
+ | Total Compression Factor | Channels | Dimensions | Modalities | Anatomies | Config File | Model File |
15
+ |----------|----------|----------|----------|----------|----------|----------|
16
+ | 16 | 1 | 2D | X-ray | Chest, Breast (FFDM) | [medvae_4x1.yaml ](model_weights/medvae_4x1.yaml)| [vae_4x_1c_2D.ckpt](model_weights/vae_4x_1c_2D.ckpt)
17
+ | 16 | 3 | 2D | X-ray | Chest, Breast (FFDM) | [medvae_4x3.yaml](model_weights/medvae_4x3.yaml) | [vae_4x_3c_2D.ckpt](model_weights/vae_4x_3c_2D.ckpt)
18
+ | 64 | 1 | 2D | X-ray | Chest, Breast (FFDM) | [medvae_8x1.yaml](model_weights/medvae_8x1.yaml) | [vae_8x_1c_2D.ckpt](model_weights/vae_8x_1c_2D.ckpt)
19
+ | 64 | 3 | 2D | X-ray | Chest, Breast (FFDM) | [medvae_8x4.yaml](model_weights/medvae_8x4.yaml) | [vae_8x_4c_2D.ckpt](model_weights/vae_8x_4c_2D.ckpt)
20
+ | 64 | 1 | 3D | MRI, CT | Whole-Body | [medvae_4x1.yaml ](model_weights/medvae_4x1.yaml) | [vae_4x_1c_3D.ckpt](model_weights/vae_4x_1c_3D.ckpt)
21
+ | 512 | 1 | 3D | MRI, CT | Whole-Body | [medvae_8x1.yaml](model_weights/vae_8x1.yaml) | [vae_8x_1c_3D.ckpt](model_weights/vae_8x_1c_3D.ckpt)
22
+
23
+ Note: Model weights and checkpoints are located in the `model_weights` folder.
24
+
25
+ ## ⚑️ Installation
26
+ To install MedVAE, you can simply run:
27
+
28
+ ```python
29
+ pip install medvae
30
+ ```
31
+
32
+ For an editable installation, use the following commands to clone and install this repository.
33
+ ```python
34
+ git clone https://github.com/StanfordMIMI/MedVAE.git
35
+ cd medvae
36
+ pip install -e .[dev]
37
+ ```
38
+
39
+
40
+ ## πŸš€ Usage Instruction
41
+
42
+ We allow for easy-to-use CLI inference tool for compressing your high-dimensional medical images into usable latents:
43
+
44
+ ```python
45
+ medvae_inference -i INPUT_FOLDER -o OUTPUT_FOLDER -model_name MED_VAE_MODEL -modality MODALITY
46
+ ```
47
+
48
+ ## πŸ“Ž Citation
49
+ If you use MedVAE, please cite the original paper:
50
+
51
+ ```bibtex
52
+ @article{varma2025medvae,
53
+ title = {MedVAE: Efficient Automated Interpretation of Medical Images with Large-Scale Generalizable Autoencoders},
54
+ author = {Maya Varma, Ashwin Kumar, Rogier van der Sluijs, Sophie Ostmeier, Louis Blankemeier, Pierre Chambon, Christian Bluethgen, Jip Prince, Curtis Langlotz, Akshay Chaudhari},
55
+ year = {2025},
56
+ publisher = {Github},
57
+ journal = {Github},
58
+ howpublished = {https://github.com/StanfordMIMI/MedVAE}
59
+ }
60
+ ```
61
+
62
+ For questions, please place a Github Issues message.