1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
|
import os
import argparse
import h5py
import numpy as np
from pathlib import Path
def test_basic_recon(reference_filename, testdata_filename):
reference_data = h5py.File(reference_filename)
input_data = h5py.File(testdata_filename)
coil_images = reference_data['dataset']['coil_images']
coil_images = np.squeeze(np.array(coil_images['real']) + 1j*np.array(coil_images['imag']))
recon_reference = np.abs(np.sqrt(np.sum(coil_images * np.conj(coil_images), 0)))
ro_length = recon_reference.shape[-1]
recon_reference = recon_reference[:, int(ro_length/4):int(ro_length/4+ro_length/2)]
# Take care of the different data types
if (len(input_data['dataset']['image_0']['data'].dtype)) == 2:
reconstruction = np.squeeze(np.array(input_data['dataset']['image_0']['data']['real']) + 1j*np.array(input_data['dataset']['image_0']['data']['imag']))
else:
reconstruction = np.squeeze(np.array(input_data['dataset']['image_0']['data']))
# We need to scale the images, since the recon does not
reconstruction = np.abs(reconstruction) / np.sqrt(coil_images.shape[-1]*coil_images.shape[-2])
assert np.linalg.norm(reconstruction - recon_reference) / np.linalg.norm(recon_reference) < 2e-5
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Validate results of an ismrmrd reconstruction')
parser.add_argument('--reference', type=str, help='Reference file', required=True)
parser.add_argument('--testdata', type=str, help='Test data file', required=True)
args = parser.parse_args()
test_basic_recon(args.reference, args.testdata)
|
