wdika commited on
Commit
039c793
·
1 Parent(s): a7f1814

Upload config

Browse files
Files changed (1) hide show
  1. readme_template.md +157 -0
readme_template.md ADDED
@@ -0,0 +1,157 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ license: apache-2.0
5
+ library_name: atommic
6
+ datasets:
7
+ - fastMRIBrainsMulticoil
8
+ thumbnail: null
9
+ tags:
10
+ - image-reconstruction
11
+ - RIM
12
+ - ATOMMIC
13
+ - pytorch
14
+ model-index:
15
+ - name: REC_RIM_fastMRIBrainsMulticoil_equispaced_4x_8x_GDCC_1_coil_NNEstimationCSM
16
+ results: []
17
+
18
+ ---
19
+
20
+
21
+ ## Model Overview
22
+
23
+ Recurrent Inference Machines (RIM) for 4x & 8x accelerated MRI Reconstruction on the fastMRIBrainsMulticoil 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/fastMRIBrainsMulticoil/conf).
38
+
39
+ ### Automatically instantiate the model
40
+
41
+ ```base
42
+ pretrained: true
43
+ checkpoint: https://huggingface.co/wdika/REC_RIM_fastMRIBrainsMulticoil_equispaced_4x_8x_GDCC_1_coil_NNEstimationCSM/blob/main/REC_RIM_fastMRIBrainsMulticoil_equispaced_4x_8x_GDCC_1_coil_NNEstimationCSM.atommic
44
+ mode: test
45
+ ```
46
+
47
+ ### Usage
48
+
49
+ You need to download the fastMRI Brains dataset to effectively use this model. Check the [fastMRIBrainsMulticoil](https://github.com/wdika/atommic/blob/main/projects/REC/fastMRIBrainsMulticoil/README.md) page for more information.
50
+
51
+
52
+ ## Model Architecture
53
+ ```base
54
+ model:
55
+ model_name: CIRIM
56
+ recurrent_layer: GRU
57
+ conv_filters:
58
+ - 64
59
+ - 64
60
+ - 2
61
+ conv_kernels:
62
+ - 5
63
+ - 3
64
+ - 3
65
+ conv_dilations:
66
+ - 1
67
+ - 2
68
+ - 1
69
+ conv_bias:
70
+ - true
71
+ - true
72
+ - false
73
+ recurrent_filters:
74
+ - 64
75
+ - 64
76
+ - 0
77
+ recurrent_kernels:
78
+ - 1
79
+ - 1
80
+ - 0
81
+ recurrent_dilations:
82
+ - 1
83
+ - 1
84
+ - 0
85
+ recurrent_bias:
86
+ - true
87
+ - true
88
+ - false
89
+ depth: 2
90
+ time_steps: 8
91
+ conv_dim: 2
92
+ num_cascades: 1
93
+ no_dc: true
94
+ keep_prediction: true
95
+ accumulate_predictions: true
96
+ dimensionality: 2
97
+ reconstruction_loss:
98
+ l1: 0.1
99
+ ssim: 0.9
100
+ estimate_coil_sensitivity_maps_with_nn: true
101
+ ```
102
+
103
+ ## Training
104
+ ```base
105
+ optim:
106
+ name: adam
107
+ lr: 1e-4
108
+ betas:
109
+ - 0.9
110
+ - 0.999
111
+ weight_decay: 0.0
112
+ sched:
113
+ name: InverseSquareRootAnnealing
114
+ min_lr: 0.0
115
+ last_epoch: -1
116
+ warmup_ratio: 0.1
117
+
118
+ trainer:
119
+ strategy: ddp_find_unused_parameters_false
120
+ accelerator: gpu
121
+ devices: 1
122
+ num_nodes: 1
123
+ max_epochs: 20
124
+ precision: 16-mixed
125
+ enable_checkpointing: false
126
+ logger: false
127
+ log_every_n_steps: 50
128
+ check_val_every_n_epoch: -1
129
+ max_steps: -1
130
+ ```
131
+
132
+ ## Performance
133
+
134
+ 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/fastMRIBrainsMulticoil/conf/targets) configuration files.
135
+
136
+ 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.
137
+
138
+ Results
139
+ -------
140
+
141
+ Evaluation against RSS targets
142
+ ------------------------------
143
+ 4x: MSE = 0.0007147 +/- 0.00289 NMSE = 0.01907 +/- 0.06354 PSNR = 33.24 +/- 6.153 SSIM = 0.8847 +/- 0.19
144
+
145
+ 8x: MSE = 0.001466 +/- 0.003407 NMSE = 0.03833 +/- 0.0846 PSNR = 29.45 +/- 5.578 SSIM = 0.8382 +/- 0.199
146
+
147
+
148
+ ## Limitations
149
+
150
+ This model was trained on the fastMRIBrainsMulticoil batch0 dataset 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.
151
+
152
+
153
+ ## References
154
+
155
+ [1] [ATOMMIC](https://github.com/wdika/atommic)
156
+
157
+ [2] Muckley MJ, Riemenschneider B, Radmanesh A, Kim S, Jeong G, Ko J, Jun Y, Shin H, Hwang D, Mostapha M, Arberet S, Nickel D, Ramzi Z, Ciuciu P, Starck JL, Teuwen J, Karkalousos D, Zhang C, Sriram A, Huang Z, Yakubova N, Lui YW, Knoll F. Results of the 2020 fastMRI Challenge for Machine Learning MR Image Reconstruction. IEEE Trans Med Imaging. 2021 Sep;40(9):2306-2317. doi: 10.1109/TMI.2021.3075856. Epub 2021 Aug 31. PMID: 33929957; PMCID: PMC8428775.