wdika commited on
Commit
7f356f2
1 Parent(s): e8476f2

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +120 -0
README.md ADDED
@@ -0,0 +1,120 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ license: apache-2.0
5
+ library_name: atommic
6
+ datasets:
7
+ - StanfordKnees2019
8
+ thumbnail: null
9
+ tags:
10
+ - image-reconstruction
11
+ - VarNet
12
+ - ATOMMIC
13
+ - pytorch
14
+ model-index:
15
+ - name: REC_VarNet_StanfordKnees2019_gaussian2d_12x_AutoEstimationCSM
16
+ results: []
17
+
18
+ ---
19
+
20
+
21
+ ## Model Overview
22
+
23
+ Variational Network (VarNet) for 12x accelerated MRI Reconstruction on the StanfordKnees2019 dataset.
24
+
25
+
26
+ ## ATOMMIC: Training
27
+
28
+ To train, fine-tune, or test the model you will need to install [ATOMMIC](https://github.com/wdika/atommic). We recommend you install it after you've installed latest Pytorch version.
29
+ ```
30
+ pip install atommic['all']
31
+ ```
32
+
33
+ ## How to Use this Model
34
+
35
+ The model is available for use in ATOMMIC, and can be used as a pre-trained checkpoint for inference or for fine-tuning on another dataset.
36
+
37
+ Corresponding configuration YAML files can be found [here](https://github.com/wdika/atommic/tree/main/projects/REC/StanfordKnees2019/conf).
38
+
39
+
40
+ ### Automatically instantiate the model
41
+
42
+ ```base
43
+ pretrained: true
44
+ checkpoint: https://huggingface.co/wdika/REC_VarNet_StanfordKnees2019_gaussian2d_12x_AutoEstimationCSM/blob/main/REC_VarNet_StanfordKnees2019_gaussian2d_12x_AutoEstimationCSM.atommic
45
+ mode: test
46
+ ```
47
+
48
+ ### Usage
49
+
50
+ You need to download the Stanford Knees 2019 dataset to effectively use this model. Check the [StanfordKnees2019](https://github.com/wdika/atommic/blob/main/projects/REC/StanfordKnees2019/README.md) page for more information.
51
+
52
+
53
+ ## Model Architecture
54
+ ```base
55
+ model:
56
+ model_name: VN
57
+ num_cascades: 8
58
+ channels: 18
59
+ pooling_layers: 4
60
+ padding_size: 11
61
+ normalize: true
62
+ no_dc: false
63
+ dimensionality: 2
64
+ reconstruction_loss:
65
+ wasserstein: 1.0
66
+ ```
67
+
68
+ ## Training
69
+ ```base
70
+ optim:
71
+ name: adamw
72
+ lr: 1e-4
73
+ betas:
74
+ - 0.9
75
+ - 0.999
76
+ weight_decay: 0.0
77
+ sched:
78
+ name: InverseSquareRootAnnealing
79
+ min_lr: 0.0
80
+ last_epoch: -1
81
+ warmup_ratio: 0.1
82
+
83
+ trainer:
84
+ strategy: ddp_find_unused_parameters_false
85
+ accelerator: gpu
86
+ devices: 1
87
+ num_nodes: 1
88
+ max_epochs: 20
89
+ precision: 16-mixed
90
+ enable_checkpointing: false
91
+ logger: false
92
+ log_every_n_steps: 50
93
+ check_val_every_n_epoch: -1
94
+ max_steps: -1
95
+ ```
96
+
97
+ ## Performance
98
+
99
+ To compute the targets using the raw k-space and the chosen coil combination method, accompanied with the chosen coil sensitivity maps estimation method, you can use [targets](https://github.com/wdika/atommic/tree/main/projects/REC/StanfordKnees2019/conf/targets) configuration files.
100
+
101
+ Evaluation can be performed using the [evaluation](https://github.com/wdika/atommic/blob/main/tools/evaluation/reconstruction.py) script for the reconstruction task, with --evaluation_type per_slice.
102
+
103
+ Results
104
+ -------
105
+
106
+ Evaluation against SENSE targets
107
+ --------------------------------
108
+ 12x: MSE = 0.001261 +/- 0.005865 NMSE = 0.04287 +/- 0.101 PSNR = 31.5 +/- 6.696 SSIM = 0.7635 +/- 0.3022
109
+
110
+
111
+ ## Limitations
112
+
113
+ This model was trained on the StanfordKnees2019 batch0 using a UNet coil sensitivity maps estimation and Geometric Decomposition Coil-Compressions to 1-coil, and might differ from the results reported on the challenge leaderboard.
114
+
115
+
116
+ ## References
117
+
118
+ [1] [ATOMMIC](https://github.com/wdika/atommic)
119
+
120
+ [2] Epperson K, Rt R, Sawyer AM, et al. Creation of Fully Sampled MR Data Repository for Compressed SENSEing of the Knee. SMRT Conference 2013;2013:1