def recon_ute(twix_file):
# recon the ute file, input the path of the Siemens twix file
## read in data
scans, evps = readTwix(twix_file)
data = np.asarray([x.data for x in scans])
# parse hdr "MEAS" for more information
meas_dict, _ = read_twix_hdr(evps[2][1])
npts = np.shape(data)[1]
nFrames = np.shape(data)[0]
gen_traj_dict = {
'npts': npts,
'nFrames': nFrames,
'traj_type': 3, #halton Spiral
'dwell_time': float(meas_dict['alDwellTime'].split()[0]) / 1000.0,
'oversampling': 3,
'ramp_time': float(meas_dict['RORampTime']),
'plat_time': 2500,
'decay_time': 60,
'del_x': 0,
'del_y': 0,
'del_z': 0,
}
x, y, z = generate_traj(**gen_traj_dict)
def vectorize(x):
return np.reshape(x, (np.prod(np.shape(x)), 1))
traj = np.squeeze(0.5 * np.stack(
(vectorize(x), vectorize(y), vectorize(z)), axis=-1))
kernel_sharpness = 0.15
kernel_extent = 7 * kernel_sharpness
recon_dict = {
'traj': traj,
'data': np.reshape(data, (npts * nFrames, 1)),
'kernel_sharpness': kernel_sharpness,
'kernel_extent': kernel_extent,
'overgrid_factor': 3,
'n_pipe_iter': 20,
'image_size': (npts, npts, npts),
'verbosity': 0,
}
print('Starting recon UTE')
uteVol = recon(**recon_dict)
uteVol = np.transpose(uteVol, (2, 1, 0))
# uteVol = complex_align(uteVol)
return (uteVol)
from GX_Twix_parser import readTwix from scipy.stats import norm import scipy.sparse as sps import numpy as np import pdb import os from GX_Recon_utils import read_twix_hdr from GX_Spec_classmap import NMR_TimeFit twix_cali_file = 'meas_005026_cali.dat' twix_dixon_file = 'meas_005026_dixon.dat' ## ************************************************part 1, fit on calibration scans, evps = readTwix(twix_cali_file) data = np.asarray([x.data for x in scans]) meas_dict = read_twix_hdr(evps[2][1]) nFids = np.shape(data)[0] nPts = np.shape(data)[1] # 200 + 1 + 20 nSkip = 100 # skip to reach steady state nGas = 1 #number of dedicated gas spectra for frequency reference nCal = 20 #number of flipangle calibration frames following the dissolved hit nDis = nFids - nCal - nGas data_dis = data[nSkip:nDis, :] data_dis_ave = np.average(data_dis, axis=0) data_gas = data[nDis, :] dwell_time = float(
def recon_dixon(twix_file):
## recon_dixon images
###################################################################
## read in data
scans, evps = readTwix(twix_file)
data_dixon = np.asarray([x.data
for x in scans[:-2]]) # the last 2 are spectrums
data_spect = np.asarray(scans[-1].data)
data_dis = data_dixon[3::2, :]
data_gas = data_dixon[
2::2, :] # the first gas is contaminated, so we throw it away
# parse hdr "MEAS" for more information
meas_dict, _ = read_twix_hdr(evps[2][1])
_, measYaps_dict = read_twix_hdr(evps[3][1]) # to read out user-set values
# reading out gradient delay
try:
gradient_delay = int(measYaps_dict['sWiPMemBlock.adFree[3]'])
except:
# if the field was not used, set it to 0
gradient_delay = 0
npts = np.shape(data_dixon)[1]
nFrames = np.shape(data_dixon)[0] + 2 # the last 2 are spectrums
TE90 = float(meas_dict['alTE'].split()[0])
gen_traj_dict = {
'npts': npts,
'nFrames': np.floor(nFrames / 2).astype(int),
'traj_type': 3, #halton Spiral
'dwell_time': float(meas_dict['alDwellTime'].split()[0]) / 1000.0,
'oversampling': 3,
'ramp_time': float(meas_dict['RORampTime']),
'plat_time': 2500,
'decay_time': 60,
'del_x': gradient_delay - 13,
'del_y': gradient_delay - 14,
'del_z': gradient_delay - 9,
}
x, y, z = generate_traj(**gen_traj_dict)
# the last 2 are used for spectroscopy
x = x[1:-1:, :]
y = y[1:-1:, :]
z = z[1:-1:, :]
def vectorize(x):
return np.reshape(x, (np.prod(np.shape(x)), 1))
data_dis, x_dis, y_dis, z_dis, nFrames_dis = remove_noise_rays(
data=data_dis, x=x, y=y, z=z, thre_snr=0.5)
print("Threw away bad dissolved FID: " +
str(nFrames / 2 - 2 - nFrames_dis))
data_gas, x_gas, y_gas, z_gas, nFrames_gas = remove_noise_rays(
data=data_gas, x=x, y=y, z=z, thre_snr=0.5)
print("Threw away bad gas FID: " + str(nFrames / 2 - 2 - nFrames_gas))
## recon gas for high SNR and high resolution
traj = np.squeeze(0.5 * np.stack(
(vectorize(x_gas), vectorize(y_gas), vectorize(z_gas)), axis=-1))
recon_dict_highSNR = {
'traj': traj,
'data': np.reshape(data_gas, (npts * nFrames_gas, 1)),
'kernel_sharpness': 0.14,
'kernel_extent': 9 * 0.14,
'overgrid_factor': 3,
'n_pipe_iter': 20,
'image_size': (npts, npts, npts),
'verbosity': 0,
}
print('Starting recon gas high SNR')
gasVol_highSNR = recon(**recon_dict_highSNR)
recon_dict_highreso = {
'traj': traj,
'data': np.reshape(data_gas, (npts * nFrames_gas, 1)),
'kernel_sharpness': 0.32,
'kernel_extent': 9 * 0.32,
'overgrid_factor': 3,
'n_pipe_iter': 20,
'image_size': (npts, npts, npts),
'verbosity': 0,
}
print('Starting recon gas high resolution')
gasVol_highreso = recon(**recon_dict_highreso)
## recon dissolved for high SNR
traj = np.squeeze(0.5 * np.stack(
(vectorize(x_dis), vectorize(y_dis), vectorize(z_dis)), axis=-1))
recon_dict_highSNR['traj'] = traj
recon_dict_highSNR['data'] = np.reshape(data_dis, (npts * nFrames_dis, 1))
print('Starting recon dissolved')
dissolvedVol = recon(**recon_dict_highSNR)
gasVol_highreso = np.transpose(gasVol_highreso, (2, 1, 0))
gasVol_highSNR = np.transpose(gasVol_highSNR, (2, 1, 0))
dissolvedVol = np.transpose(dissolvedVol, (2, 1, 0))
return gasVol_highSNR, gasVol_highreso, dissolvedVol, TE90
def spect_fit(twix_cali_file, twix_dixon_file, Subject_ID):
## spectroscopic fitting on calibration file and dixon bonus spectrum
## ************************************************part 1, fit on calibration
scans, evps = readTwix(twix_cali_file)
data = np.asarray([x.data for x in scans])
meas_dict,_ = read_twix_hdr(evps[2][1])
nFids = np.shape(data)[0]
nPts = np.shape(data)[1]
# 200 + 1 + 20
nSkip = 100 # skip to reach steady state
nGas = 1 #number of dedicated gas spectra for frequency reference
nCal = 20 #number of flipangle calibration frames following the dissolved hit
nDis = nFids - nCal - nGas
data_dis = data[nSkip:nDis,:]
data_dis_ave = np.average(data_dis,axis=0)
data_gas = data[nDis,:]
dwell_time = float(meas_dict['alDwellTime'].split()[0])*1e-9 # unit in second
t = np.array(range(0,nPts))*dwell_time
## initial fit from the calibration to determine frequency and fwhm of gas and dissolved
gasfit = NMR_TimeFit(time_signal=data_gas, t=t, area= 1e-4, freq=-84,
fwhm=30, phase=0, line_boardening=0,zeropad_size=10000,method='lorenzian')
gasfit.fit_time_signal()
disfit = NMR_TimeFit(time_signal=data_dis_ave, t=t, area=[1,1,1],freq=[0,-700,-7400],
fwhmL=[250,200,30],fwhmG=[0,200,0],phase=[0,0,0],line_boardening=0,zeropad_size=np.size(t),method='voigt')
lb = np.stack(([-np.inf,-np.inf,-np.inf],[-100,-900,-np.inf],[-np.inf,-np.inf,-np.inf],[-np.inf,-np.inf,-np.inf],[-np.inf,-np.inf,-np.inf])).flatten()
ub = np.stack(([+np.inf,+np.inf,+np.inf],[+100,-500,+np.inf],[+np.inf,+np.inf,+np.inf],[+np.inf,+np.inf,+np.inf],[+np.inf,+np.inf,+np.inf])).flatten()
bounds = (lb,ub)
disfit.fit_time_signal(bounds)
# pdb.set_trace()
## ************************************************part 2, fit on dixon bonus
scans, evps = readTwix(twix_dixon_file)
data_dixon = np.asarray(scans[-1].data)
meas_dict,_ = read_twix_hdr(evps[2][1])
nPts = np.size(data_dixon)
dwell_time = float(meas_dict['alDwellTime'].split()[0])*1e-9 # unit in second
t = np.array(range(0,nPts))*dwell_time
dixonfit = NMR_TimeFit(time_signal=data_dixon, t=t, area=[1,1,1],freq=[0,-700,-7400],
fwhmL=[250,200,30],fwhmG=[0,200,0],phase=[0,0,0],line_boardening=0,zeropad_size=np.size(t),method='voigt')
dixonfit.fit_time_signal()
# fit again with the freq and fwhm from the calibration fitting
area = dixonfit.area
freq = disfit.freq - gasfit.freq
fwhmL = disfit.fwhmL
fwhmG = disfit.fwhmG
phase = dixonfit.phase
fwhmL[fwhmL<0] = 1e-4
fwhmG[fwhmG<0] = 1e-4
assert np.prod(fwhmL>0), "There is a negative value in Dissolved fitting fwhmL"
assert np.prod(fwhmG>0), "There is a negative value in Dissolved fitting fwhmG"
# set up bounds to constrain frequency and fwhm change
lb = np.stack((0.33*area,freq-1.0,0.9*fwhmL,0.9*fwhmG-1.0,[-np.inf,-np.inf,-np.inf])).flatten()
ub = np.stack((3.00*area,freq+1.0,1.1*fwhmL,1.1*fwhmG+1.0,[+np.inf,+np.inf,+np.inf])).flatten()
# lb = np.stack((0.33*area,[-np.inf,-np.inf,-np.inf],0.9*fwhmL,0.9*fwhmG-1.0,[-np.inf,-np.inf,-np.inf])).flatten()
# ub = np.stack((3.00*area,[+np.inf,+np.inf,+np.inf],1.1*fwhmL,1.1*fwhmG+1.0,[+np.inf,+np.inf,+np.inf])).flatten()
bounds = (lb,ub)
dixonfit = NMR_TimeFit(time_signal=data_dixon, t=t, area=area,freq=freq,
fwhmL=fwhmL,fwhmG=fwhmG,phase=phase,line_boardening=0,zeropad_size=np.size(t),method='voigt')
dixonfit.fit_time_signal(bounds)
fit_box_cali = {
'Area': disfit.area,
'Freq': disfit.freq,
'FwhmL': disfit.fwhmL,
'FwhmG': disfit.fwhmG,
'Phase': disfit.phase,
}
print_fit(fit_box = fit_box_cali, Subject_ID = Subject_ID, key = 1)
fit_box = {
'Area': dixonfit.area,
'Freq': dixonfit.freq,
'FwhmL': dixonfit.fwhmL,
'FwhmG': dixonfit.fwhmG,
'Phase': dixonfit.phase,
}
print_fit(fit_box = fit_box, Subject_ID = Subject_ID, key =2)
# pdb.set_trace()
RBC2barrier = dixonfit.area[0]/dixonfit.area[1]
return RBC2barrier, fit_box
def spect_calibration(twix_cali_file, result_file_path):
## ************************************************part 1, fit on calibration
scans, evps = readTwix(twix_cali_file)
data = np.asarray([x.data for x in scans])
meas_dict, _ = read_twix_hdr(evps[2][1])
dicom_dict, _ = read_twix_hdr(evps[1][1])
# fetch useful values
nFids = np.shape(data)[0]
nPts = np.shape(data)[1]
dwell_time = float(
meas_dict['alDwellTime'].split()[0]) * 1e-9 # unit in second
t = np.array(range(0, nPts)) * dwell_time
te = float(meas_dict['alTE'].split()[0])
freq = int(dicom_dict['lFrequency'])
# 200 + 1 + 20
nSkip = 100 # skip to reach steady state
nGas = 1 #number of dedicated gas spectra for frequency reference
nCal = 20 #number of flipangle calibration frames following the dissolved hit
nDis = nFids - nGas - nCal
# some parameters that may be useful
freq_std = 34091550
freq_tol = 200
deltaPhase1_tol = 90
TrueRefScale_tol = 1.17
SNR_tol = 25
flip_angle_target = 20
rbc_bar_adjust = 80.0
# split gas and dissolved data
nDis = nFids - nCal - nGas
data_dis = data[nSkip:nDis, :]
data_dis_ave = np.average(data_dis, axis=0)
data_gas = data[nDis, :]
## initial fit from the calibration to determine frequency and fwhm of gas and dissolved
gasfit = NMR_TimeFit(time_signal=data_gas,
t=t,
area=1e-4,
freq=-84,
fwhm=30,
phase=0,
line_boardening=0,
zeropad_size=10000,
method='lorenzian')
gasfit.fit_time_signal()
disfit = NMR_TimeFit(time_signal=data_dis_ave,
t=t,
area=[1, 1, 1],
freq=[0, -700, -7400],
fwhmL=[250, 200, 30],
fwhmG=[0, 200, 0],
phase=[0, 0, 0],
line_boardening=0,
zeropad_size=np.size(t),
method='voigt')
disfit.fit_time_signal()
# 1 . report frequency
freq_target = freq + gasfit.freq
freq_target = int(np.asscalar(np.round(freq_target)))
stream = '129Xe cloud is delivering calibration results:\r\n'
print('\r\nFrequency_target = {:8.0f} Hz'.format(freq_target))
stream = stream + 'Frequency_target = {:8.0f} Hz *****\r\n'.format(
freq_target)
if ((freq_target - freq_std) > freq_tol):
print('***Warning! Frequency adjust exceeds tolerances; Check system')
stream = stream + '***Warning! Frequency adjust exceeds tolerances; Check system\r\n'
# 2. report TE90
deltaPhase = disfit.phase[1] - disfit.phase[0]
deltaPhase = np.mod(abs(deltaPhase), 180)
deltaFreq = abs(disfit.freq[1] - disfit.freq[0])
deltaTE90 = (90 - deltaPhase) / (360 * deltaFreq)
TE90 = te + deltaTE90 * 1e6 # in usec
print("TE90 = {:3.2f} ms".format(TE90 / 1000))
stream = stream + "TE90 = {:3.2f} ms *****\r\n".format(TE90 / 1000)
if abs(deltaPhase) > 90:
print('***WARNING! Phi_cal = {:3.0f}{}; Use min TE!'.format(
deltaPhase, u'\u00b0'.encode('utf8')))
stream = stream + '***WARNING! Phi_cal = {:3.0f}{}; Use min TE!\r\n'.format(
deltaPhase, u'\u00b0'.encode('utf8'))
# 3. report reference voltage (Flip angle)
calData = data[(nDis + 1):(nDis + nCal + 1), :]
flipCalAmps = np.amax(abs(calData), axis=1)
guess = [np.max(flipCalAmps), 20.0 * np.pi / 180]
x_data = np.array(range(1, len(flipCalAmps) + 1))
y_data = flipCalAmps
# curve fitting using trust region reflection algorithm
def calc_fitting_residual(coef):
y_fit = coef[0] * np.cos(coef[1])**(x_data - 1)
residual = (y_fit - y_data).flatten()
return residual
max_nfev = 13000
bounds = ([-np.inf, -np.inf], [np.inf, np.inf])
fit_result = least_squares(fun=calc_fitting_residual,
x0=guess,
jac='2-point',
bounds=bounds,
method='dogbox',
max_nfev=max_nfev)
flip_angle = abs(fit_result['x'][1] * 180 / np.pi)
v_ref_scale_factor = flip_angle_target / flip_angle
print('True Ref scale factor = {:3.3f}'.format(v_ref_scale_factor))
stream = stream + 'True Ref scale factor = {:3.3f} *****\r\n'.format(
v_ref_scale_factor)
print('For 600V calibration, True_Ref = {:3.0f} V'.format(
600 * v_ref_scale_factor))
stream = stream + 'For 600V calibration, True_Ref = {:3.0f} V *****\r\n'.format(
600 * v_ref_scale_factor)
if (v_ref_scale_factor > TrueRefScale_tol):
print('***Warning! Excessive calibration scale factor; check system')
stream = stream + '***Warning! Excessive calibration scale factor; check system\r\n'
# 4. calculate and report SNR for dissolved peaks
time_fit = disfit.calc_time_sig(disfit.t)
time_res = time_fit - disfit.time_signal
n25pct = int(round(len(time_res) / 4.0))
std25 = np.std(time_res[-n25pct:])
SNR_dis = disfit.area / std25
print('\r\nSNR for dissolved peaks = {:3.1f}, {:3.1f}, {:3.1f}'.format(
SNR_dis[0], SNR_dis[1], SNR_dis[2]))
stream = stream + '\r\nSNR for dissolved peaks = {:3.1f}, {:3.1f}, {:3.1f}\r\n'.format(
SNR_dis[0], SNR_dis[1], SNR_dis[2])
# 5. calculate and report SNR for gas
time_fit = gasfit.calc_time_sig(gasfit.t)
time_res = time_fit - gasfit.time_signal
n25pct = int(round(len(time_res) / 4.0))
std25 = np.std(time_res[-n25pct:])
SNR_gas = gasfit.area / std25
print('SNR for gas peak = {:3.1f}'.format(SNR_gas[0]))
stream = stream + 'SNR for gas peak = {:3.1f}\r\n'.format(SNR_gas[0])
if SNR_gas < SNR_tol:
print('***WARNING! Gas FID SNR below minimums; Check Coil Plug')
stream = stream + '***WARNING! Gas FID SNR below minimums; Check Coil Plug\r\n'
# 6. Quantify ammount of off resonance excitation
gas_dis_ratio = disfit.area[2] / sum(disfit.area[:2])
print('\r\ngas_dis_ratio = {:3.3f}'.format(gas_dis_ratio))
stream = stream + '\r\ngas_dis_ratio = {:3.3f}\r\n'.format(gas_dis_ratio)
# 7. Quantify RBC:Barrier ratio
rbc_bar_ratio = disfit.area[0] / disfit.area[1]
print('rbc_bar_ratio = {:3.3f}'.format(rbc_bar_ratio))
print('RbcBar_{:2.0f} = {:3.3f}'.format(
rbc_bar_adjust, rbc_bar_ratio * rbc_bar_adjust / 100.0))
stream = stream + 'rbc_bar_ratio = {:3.3f}\r\n'.format(rbc_bar_ratio)
stream = stream + 'RbcBar_{:2.0f} = {:3.3f}\r\n'.format(
rbc_bar_adjust, rbc_bar_ratio * rbc_bar_adjust / 100.0)
with open(result_file_path, 'w') as cali_doc:
cali_doc.write(stream)
return stream