def test_fit_FSLModel_on_invivo_sim():
FIDs, hdr, trueconcs = syntheticFromBasisFile(basis_path,
noisecovariance=[[1E-3]],
broadening=(9.0, 9.0),
concentrations={'Mac': 2.0})
basis, names, header = mrs_io.read_basis(basis_path)
mrs = MRS(FID=FIDs,
header=hdr,
basis=basis,
basis_hdr=header[0],
names=names)
mrs.processForFitting()
metab_groups = [0] * mrs.numBasis
Fitargs = {
'ppmlim': [0.2, 4.2],
'method': 'MH',
'baseline_order': -1,
'metab_groups': metab_groups,
'MHSamples': 100
}
res = fit_FSLModel(mrs, **Fitargs)
fittedRelconcs = res.getConc(scaling='internal', metab=mrs.names)
answers = np.asarray(trueconcs)
answers /= (answers[names.index('Cr')] + trueconcs[names.index('PCr')])
assert np.allclose(fittedRelconcs, answers, atol=5E-2)
def test_quantifyWater():
basis, names, headerb = mrsio.read_basis(basisfile)
crIndex = names.index('Cr')
data, header = mrsio.read_FID(metabfile)
dataw, headerw = mrsio.read_FID(h2ofile)
mrs = MRS(FID=data,
header=header,
basis=basis[:, crIndex],
names=['Cr'],
basis_hdr=headerb[crIndex],
H2O=dataw)
mrs.check_FID(repair=True)
mrs.check_Basis(repair=True)
Fitargs = {
'ppmlim': [0.2, 5.2],
'method': 'MH',
'baseline_order': -1,
'metab_groups': [0]
}
res = fit_FSLModel(mrs, **Fitargs)
tissueFractions = {'GM': 0.6, 'WM': 0.4, 'CSF': 0.0}
TE = 0.03
T2dict = {
'H2O_GM': 0.110,
'H2O_WM': 0.080,
'H2O_CSF': 2.55,
'METAB': 0.160
}
res.calculateConcScaling(mrs,
referenceMetab=['Cr'],
waterRefFID=mrs.H2O,
tissueFractions=tissueFractions,
TE=TE,
T2=T2dict,
waterReferenceMetab='Cr',
wRefMetabProtons=5,
reflimits=(2, 5),
verbose=False)
print(res.getConc(scaling='raw'))
print(res.getConc(scaling='internal'))
print(res.getConc(scaling='molality'))
print(res.getConc(scaling='molarity'))
assert np.allclose(res.getConc(scaling='internal'), 1.0)
assert np.allclose(res.getConc(scaling='molarity'), 10.59, atol=1E-1)
def data():
noiseCov = 0.001
amplitude = np.asarray([0.5, 0.5, 1.0]) * 10
chemshift = np.asarray([3.0, 3.05, 2.0]) - 4.65
lw = [10, 10, 10]
phases = [0, 0, 0]
g = [0, 0, 0]
basisNames = ['Cr', 'PCr', 'NAA']
begintime = 0.00005
basisFIDs = []
for idx, _ in enumerate(amplitude):
tmp, basisHdr = syntheticFID(noisecovariance=[[0.0]],
chemicalshift=[chemshift[idx] + 0.1],
amplitude=[1.0],
linewidth=[lw[idx] / 5],
phase=[phases[idx]],
g=[g[idx]],
begintime=0)
basisFIDs.append(tmp[0])
basisFIDs = np.asarray(basisFIDs)
synFID, synHdr = syntheticFID(noisecovariance=[[noiseCov]],
chemicalshift=chemshift,
amplitude=amplitude,
linewidth=lw,
phase=phases,
g=g,
begintime=begintime)
synMRS = MRS(FID=synFID[0],
header=synHdr,
basis=basisFIDs,
basis_hdr=basisHdr,
names=basisNames)
metab_groups = [0] * synMRS.numBasis
Fitargs = {
'ppmlim': [0.2, 4.2],
'method': 'MH',
'baseline_order': -1,
'metab_groups': metab_groups,
'MHSamples': 100,
'disable_mh_priors': True
}
res = fit_FSLModel(synMRS, **Fitargs)
return res, amplitude
def test_fit_FSLModel_Newton(data):
mrs = data[0]
amplitudes = data[1]
metab_groups = [0] * mrs.numBasis
Fitargs = {
'ppmlim': [0.2, 4.2],
'method': 'Newton',
'baseline_order': -1,
'metab_groups': metab_groups
}
res = fit_FSLModel(mrs, **Fitargs)
fittedconcs = res.getConc(metab=mrs.names)
fittedRelconcs = res.getConc(scaling='internal', metab=mrs.names)
assert np.allclose(fittedconcs, amplitudes, atol=1E-1)
assert np.allclose(fittedRelconcs,
amplitudes / (amplitudes[0] + amplitudes[1]),
atol=1E-1)
def test_calcQC():
# Syntetic data
synFID, synHdr = syntheticFID(noisecovariance=[[0.1]],
points=2 * 2048,
chemicalshift=[0],
amplitude=[6.0],
linewidth=[10])
synFIDNoise, synHdrNoise = syntheticFID(noisecovariance=[[0.1]],
points=2 * 2048,
chemicalshift=[0],
amplitude=[0],
linewidth=[10])
basisFID, basisHdr = syntheticFID(noisecovariance=[[0.0]],
points=2 * 2048,
chemicalshift=[0],
amplitude=[0.1],
linewidth=[2])
synMRS = MRS(FID=synFID[0], header=synHdr)
synMRSNoise = MRS(FID=synFIDNoise[0], header=synHdrNoise)
synMRSNoNoise = MRS(FID=synHdr['noiseless'], header=synHdr)
synMRS_basis = MRS(FID=synFID[0],
header=synHdr,
basis=basisFID[0],
basis_hdr=basisHdr,
names=['Peak1'])
truenoiseSD = np.sqrt(synHdrNoise['cov'][0, 0])
pureNoiseMeasured = np.std(synMRSNoise.getSpectrum())
realnoise = np.std(np.real(synMRSNoise.getSpectrum()))
imagNoise = np.std(np.imag(synMRSNoise.getSpectrum()))
print(
f'True cmplx noise = {truenoiseSD:0.3f}, pure noise measured = {pureNoiseMeasured:0.3f} (real/imag = {realnoise:0.3f}/{imagNoise:0.3f})'
)
# Calc SNR without apodisation from the no noise and pure noise spectra
truePeakHeight = np.max(np.real(synMRSNoNoise.getSpectrum()))
SNR_noApod = truePeakHeight / pureNoiseMeasured
print(
f'SNR no apod: {SNR_noApod:0.1f} ({truePeakHeight:0.2e}/{pureNoiseMeasured:0.2e})'
)
# Calc SNR with apodisation from the no noise and pure noise spectra
trueLW = synHdr['inputopts']['linewidth'][0]
trueApodSpec_Noise = specApodise(synMRSNoise, trueLW)
apodNoise = np.std(trueApodSpec_Noise)
trueApodSpec_noNoise = specApodise(synMRSNoNoise, trueLW)
peakHeigtApod = np.max(np.real(trueApodSpec_noNoise))
SNR = peakHeigtApod / apodNoise
print(f'SNR w. apod: {SNR:0.1f} ({peakHeigtApod:0.2e}/{apodNoise:0.2e})')
metab_groups = [0]
Fitargs = {
'ppmlim': [2.65, 6.65],
'method': 'Newton',
'baseline_order': -1,
'metab_groups': [0]
}
res = fit_FSLModel(synMRS_basis, **Fitargs)
fwhm_test, SNRObj = calcQC(synMRS_basis, res, ppmlim=[2.65, 6.65])
print(f'Measured FWHM: {fwhm_test.mean().to_numpy()[0]:0.1f}')
print(f'Measured spec SNR: {SNRObj.spectrum:0.1f}')
print(f'Measured peak SNR: {SNRObj.peaks.mean().to_numpy()[0]:0.1f}')
assert np.isclose(fwhm_test.mean().to_numpy(), trueLW, atol=1E0)
assert np.isclose(SNRObj.spectrum, SNR_noApod, atol=1E1)
assert np.isclose(SNRObj.peaks.mean().to_numpy(), SNR, atol=2E1)