recotwix comprises a set of utilities for reconstructing MR images from Siemens rawdata files (.dat). The goal is to offer distinct Python scripts (classes) for reconstructing various sequence types. These classes inherit from 'recotwix', the parent class, which furnishes fundamental reconstruction routines and several general utilities.
recotwix uses twixtools for raw data reading purposes. It has undergone testing and validation with Python 3.10. Other packages needed:
- numpy
- torch
- scipy
- nibabel
- tqdm
recotwix also employs BART for coil sensitivity calculation and combination.
open terminal and simply navigate to the recotwix and install it with pip:
pip install .
The provided examples showcase the reconstruction of a B0 map derived from a double-echo GRE sequence and a B1+ map from a 3D satTFL sequence.
%matplotlib widget
from recoB0 import recoB0
from pytools import plot
filename = '/DATA2/rawdata/twixtools_samples/meas_MID00575_FID29449_aa_B0Phantom.dat'
twix = recoB0(filename)
plot.plot3D(twix.img_b0, title='B0 map (wrapped)')
plot.plot3D(twix.get_b0hz(), clim = [-300, 300], title='B0 map (unwrapped, Hz)')Software version: VD/VE (!?)
Scan 0
100%|██████████| 949M/949M [00:01<00:00, 672MB/s]
Fourier transform: 100%|██████████| 32/32 [00:00<00:00, 101.98it/s]
Calculating B0 map. ΔTE = 3500 μs
Unwrapping B0...
%matplotlib widget
from recoB1TFLpTx import recoB1TFLpTx
from pytools import plot
import torch
filename = '/DATA2/rawdata/twixtools_samples/meas_MID00866_FID61816_db_TFLb1_ICE_48Slc.dat'
twix = recoB1TFLpTx(filename)
print(twix.img_cp.shape)
print(twix.img.shape)
print(twix.img.abs().index_select(twix.dim_info['Rep']['ind'], torch.Tensor([0]).int()).shape)
plot.plot3D(twix.img_cp.abs(), title='CP image')
plot.plot3D(twix.img_fa.abs().index_select(twix.dim_info['Rep']['ind'], torch.Tensor([12]).int()), title='B1+ magnitude, Tx 13')Software version: VD/VE (!?)
Scan 0
100%|██████████| 513M/513M [00:01<00:00, 364MB/s]
kspace shape: torch.Size([1, 1, 1, 1, 1, 1, 1, 34, 1, 1, 48, 1, 1, 64, 32, 48])
kspace corrected shape: torch.Size([1, 1, 1, 1, 1, 1, 1, 34, 1, 1, 48, 1, 1, 64, 32, 64])
POCS reconstruction along dim = 15 started...
Fourier transform: 100%|██████████| 32/32 [00:00<00:00, 233.13it/s]
Fourier transform: 100%|██████████| 32/32 [00:00<00:00, 223.96it/s]
Fourier transform: 100%|██████████| 32/32 [00:00<00:00, 436.70it/s]
Fourier transform: 100%|██████████| 32/32 [00:00<00:00, 370.96it/s]
Fourier transform: 100%|██████████| 32/32 [00:00<00:00, 337.88it/s]
Fourier transform: 100%|██████████| 32/32 [00:00<00:00, 399.63it/s]
Fourier transform: 100%|██████████| 32/32 [00:00<00:00, 345.54it/s]
Fourier transform: 100%|██████████| 32/32 [00:00<00:00, 401.82it/s]
Fourier transform: 100%|██████████| 32/32 [00:00<00:00, 341.23it/s]
Fourier transform: 100%|██████████| 32/32 [00:00<00:00, 403.41it/s]
Fourier transform: 100%|██████████| 32/32 [00:00<00:00, 342.44it/s]
Fourier transform: 100%|██████████| 32/32 [00:00<00:00, 377.41it/s]
Fourier transform: 100%|██████████| 32/32 [00:00<00:00, 231.85it/s]
Done!
Computing coil sensitivity...
kernel_size: 32
Calibration region 32x24x24
Done.
Done!
Combining coils...
Fourier transform: 100%|██████████| 32/32 [00:01<00:00, 17.91it/s]
GPU reconstruction
Size: 6684672 Samples: 1671167 Acc: 4.00
Inverse scaling of the data: 0.000001
l2 regularization: 0.000100
Regularization terms: 1, Supporting variables: 0
conjugate gradients
lsqr: add GPU wrapper
cg: 30
Total Time: 7.485684
Done!
torch.Size([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 48, 1, 1, 64, 1, 64])
torch.Size([1, 1, 1, 1, 1, 1, 1, 34, 1, 1, 48, 1, 1, 64, 1, 64])
torch.Size([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 48, 1, 1, 64, 1, 64])
