def test_increment():
# `example_mod` is defined in ../cuvec/src/example_mod/
from cuvec.example_mod import increment2d_f
a = cu.zeros((1337, 42), 'f')
assert (a == 0).all()
res = cu.asarray(increment2d_f(a.cuvec, a.cuvec))
assert (a == 1).all()
assert (res == 1).all()
a[:] = 0
assert (a == 0).all()
assert (res == 0).all()
res = cu.asarray(increment2d_f(a))
assert (res == 1).all()
def test_increment_return():
from cuvec.example_mod import increment2d_f
a = cu.zeros((1337, 42), 'f')
assert (a == 0).all()
res = cu.asarray(increment2d_f(a, a))
assert (a == 1).all()
del a
assert (res == 1).all()
def test_np_types():
from cuvec.example_mod import increment2d_f
f = cu.zeros((1337, 42), 'f')
d = cu.zeros((1337, 42), 'd')
cu.asarray(increment2d_f(f))
cu.asarray(increment2d_f(f, f))
with raises(TypeError):
cu.asarray(increment2d_f(d))
with raises(SystemError):
# the TypeError is suppressed since a new output is generated
cu.asarray(increment2d_f(f, d))
def test_asarray():
v = cu.asarray(np.random.random(shape))
w = cu.CuVec(v)
assert w.cuvec == v.cuvec
assert (w == v).all()
assert np.asarray(w.cuvec).data == np.asarray(v.cuvec).data
x = cu.asarray(w.cuvec)
assert x.cuvec == v.cuvec
assert (x == v).all()
assert np.asarray(x.cuvec).data == np.asarray(v.cuvec).data
y = cu.asarray(x.tolist())
assert y.cuvec != v.cuvec
assert (y == v).all()
assert np.asarray(y.cuvec).data == np.asarray(v.cuvec).data
z = cu.asarray(v[:])
assert z.cuvec != v.cuvec
assert (z == v[:]).all()
assert np.asarray(z.cuvec).data == np.asarray(v.cuvec).data
s = cu.asarray(v[1:])
assert s.cuvec != v.cuvec
assert (s == v[1:]).all()
assert np.asarray(s.cuvec).data != np.asarray(v.cuvec).data
def test_cuda_array_interface():
cupy = importorskip("cupy")
v = cu.asarray(np.random.random(shape))
assert hasattr(v, '__cuda_array_interface__')
c = cupy.asarray(v)
assert (c == v).all()
c[0, 0, 0] = 1
cu.dev_sync()
assert c[0, 0, 0] == v[0, 0, 0]
c[0, 0, 0] = 0
cu.dev_sync()
assert c[0, 0, 0] == v[0, 0, 0]
ndarr = v + 1
assert ndarr.shape == v.shape
assert ndarr.dtype == v.dtype
with raises(AttributeError):
ndarr.__cuda_array_interface__
def osemone(datain,
mumaps,
hst,
scanner_params,
recmod=3,
itr=4,
fwhm=0.,
psf=None,
mask_radius=29.,
decay_ref_time=None,
attnsino=None,
sctsino=None,
randsino=None,
normcomp=None,
emmskS=False,
frmno='',
fcomment='',
outpath=None,
fout=None,
store_img=False,
store_itr=None,
ret_sinos=False):
'''
OSEM image reconstruction with several modes
(with/without scatter and/or attenuation correction)
Args:
psf: Reconstruction with PSF, passed to `psf_config`
'''
# > Get particular scanner parameters: Constants, transaxial and axial LUTs
Cnt = scanner_params['Cnt']
txLUT = scanner_params['txLUT']
axLUT = scanner_params['axLUT']
# ---------- sort out OUTPUT ------------
# -output file name for the reconstructed image
if outpath is None:
opth = os.path.join(datain['corepath'], 'reconstructed')
else:
opth = outpath
# > file output name (the path is ignored if given)
if fout is not None:
# > get rid of folders
fout = os.path.basename(fout)
# > get rid of extension
fout = fout.split('.')[0]
if store_img is True or store_itr is not None:
mmraux.create_dir(opth)
return_ssrb, return_mask = ret_sinos, ret_sinos
# ----------
log.info('reconstruction in mode: %d', recmod)
# get object and hardware mu-maps
muh, muo = mumaps
# get the GPU version of the image dims
mus = mmrimg.convert2dev(muo + muh, Cnt)
# remove gaps from the prompt sino
psng = mmraux.remgaps(hst['psino'], txLUT, Cnt)
# ========================================================================
# GET NORM
# -------------------------------------------------------------------------
if normcomp is None:
ncmp, _ = mmrnorm.get_components(datain, Cnt)
else:
ncmp = normcomp
log.warning('using user-defined normalisation components')
nsng = mmrnorm.get_norm_sino(datain,
scanner_params,
hst,
normcomp=ncmp,
gpu_dim=True)
# ========================================================================
# ========================================================================
# ATTENUATION FACTORS FOR COMBINED OBJECT AND BED MU-MAP
# -------------------------------------------------------------------------
# > combine attenuation and norm together depending on reconstruction mode
if recmod == 0:
asng = np.ones(psng.shape, dtype=np.float32)
else:
# > check if the attenuation sino is given as an array
if isinstance(attnsino, np.ndarray) \
and attnsino.shape==(Cnt['NSN11'], Cnt['NSANGLES'], Cnt['NSBINS']):
asng = mmraux.remgaps(attnsino, txLUT, Cnt)
log.info('using provided attenuation factor sinogram')
elif isinstance(attnsino, np.ndarray) \
and attnsino.shape==(Cnt['Naw'], Cnt['NSN11']):
asng = attnsino
log.info('using provided attenuation factor sinogram')
else:
asng = cu.zeros(psng.shape, dtype=np.float32)
petprj.fprj(asng.cuvec,
cu.asarray(mus).cuvec, txLUT, axLUT,
np.array([-1], dtype=np.int32), Cnt, 1)
# > combine attenuation and normalisation
ansng = asng * nsng
# ========================================================================
# ========================================================================
# Randoms
# -------------------------------------------------------------------------
if isinstance(randsino, np.ndarray) \
and randsino.shape==(Cnt['NSN11'], Cnt['NSANGLES'], Cnt['NSBINS']):
rsino = randsino
rsng = mmraux.remgaps(randsino, txLUT, Cnt)
else:
rsino, snglmap = randoms(hst, scanner_params)
rsng = mmraux.remgaps(rsino, txLUT, Cnt)
# ========================================================================
# ========================================================================
# SCAT
# -------------------------------------------------------------------------
if recmod == 2:
if sctsino is not None:
ssng = mmraux.remgaps(sctsino, txLUT, Cnt)
elif sctsino is None and os.path.isfile(datain['em_crr']):
emd = nimpa.getnii(datain['em_crr'])
ssn = vsm(
datain,
mumaps,
emd['im'],
scanner_params,
histo=hst,
rsino=rsino,
prcnt_scl=0.1,
emmsk=False,
)
ssng = mmraux.remgaps(ssn, txLUT, Cnt)
else:
raise ValueError(
"No emission image available for scatter estimation! " +
" Check if it's present or the path is correct.")
else:
ssng = np.zeros(rsng.shape, dtype=rsng.dtype)
# ========================================================================
log.info('------ OSEM (%d) -------', itr)
# ------------------------------------
Sn = 14 # number of subsets
# -get one subset to get number of projection bins in a subset
Sprj, s = get_subsets14(0, scanner_params)
Nprj = len(Sprj)
# -init subset array and sensitivity image for a given subset
sinoTIdx = np.zeros((Sn, Nprj + 1), dtype=np.int32)
# -init sensitivity images for each subset
imgsens = np.zeros((Sn, Cnt['SZ_IMY'], Cnt['SZ_IMX'], Cnt['SZ_IMZ']),
dtype=np.float32)
tmpsens = cu.zeros((Cnt['SZ_IMY'], Cnt['SZ_IMX'], Cnt['SZ_IMZ']),
dtype=np.float32)
for n in range(Sn):
# first number of projection for the given subset
sinoTIdx[n, 0] = Nprj
sinoTIdx[n, 1:], s = get_subsets14(n, scanner_params)
# sensitivity image
petprj.bprj(tmpsens.cuvec,
cu.asarray(ansng[sinoTIdx[n, 1:], :]).cuvec, txLUT, axLUT,
sinoTIdx[n, 1:], Cnt)
imgsens[n] = tmpsens
del tmpsens
# -------------------------------------
# -mask for reconstructed image. anything outside it is set to zero
msk = mmrimg.get_cylinder(
Cnt, rad=mask_radius, xo=0, yo=0, unival=1, gpu_dim=True) > 0.9
# -init image
img = np.ones((Cnt['SZ_IMY'], Cnt['SZ_IMX'], Cnt['SZ_IMZ']),
dtype=np.float32)
# -decay correction
lmbd = np.log(2) / resources.riLUT[Cnt['ISOTOPE']]['thalf']
if Cnt['DCYCRR'] and 't0' in hst and 'dur' in hst:
# > decay correct to the reference time (e.g., injection time) if provided
# > otherwise correct in reference to the scan start time (using the time
# > past from the start to the start time frame)
if decay_ref_time is not None:
tref = decay_ref_time
else:
tref = hst['t0']
dcycrr = np.exp(
lmbd * tref) * lmbd * hst['dur'] / (1 - np.exp(-lmbd * hst['dur']))
# apply quantitative correction to the image
qf = ncmp['qf'] / resources.riLUT[Cnt['ISOTOPE']]['BF'] / float(
hst['dur'])
qf_loc = ncmp['qf_loc']
elif not Cnt['DCYCRR'] and 't0' in hst and 'dur' in hst:
dcycrr = 1.
# apply quantitative correction to the image
qf = ncmp['qf'] / resources.riLUT[Cnt['ISOTOPE']]['BF'] / float(
hst['dur'])
qf_loc = ncmp['qf_loc']
else:
dcycrr = 1.
qf = 1.
qf_loc = 1.
# -affine matrix for the reconstructed images
B = mmrimg.image_affine(datain, Cnt)
# resolution modelling
psfkernel = psf_config(psf, Cnt)
# -time it
stime = time.time()
# import pdb; pdb.set_trace()
# ========================================================================
# OSEM RECONSTRUCTION
# -------------------------------------------------------------------------
with trange(itr,
desc="OSEM",
disable=log.getEffectiveLevel() > logging.INFO,
leave=log.getEffectiveLevel() <= logging.INFO) as pbar:
for k in pbar:
petprj.osem(img, psng, rsng, ssng, nsng, asng, sinoTIdx, imgsens,
msk, psfkernel, txLUT, axLUT, Cnt)
if np.nansum(img) < 0.1:
log.warning(
'it seems there is not enough true data to render reasonable image'
)
# img[:]=0
itr = k
break
if recmod >= 3 and k < itr - 1 and itr > 1:
sct_time = time.time()
sct = vsm(datain,
mumaps,
mmrimg.convert2e7(img, Cnt),
scanner_params,
histo=hst,
rsino=rsino,
emmsk=emmskS,
return_ssrb=return_ssrb,
return_mask=return_mask)
if isinstance(sct, dict):
ssn = sct['sino']
else:
ssn = sct
ssng = mmraux.remgaps(ssn, txLUT, Cnt)
pbar.set_postfix(scatter="%.3gs" % (time.time() - sct_time))
# save images during reconstruction if requested
if store_itr and (k + 1) in store_itr:
im = mmrimg.convert2e7(img * (dcycrr * qf * qf_loc), Cnt)
if fout is None:
fpet = os.path.join(opth, (
os.path.basename(datain['lm_bf'])[:16].replace(
'.', '-') +
f"{frmno}_t{hst['t0']}-{hst['t1']}sec_itr{k+1}{fcomment}_inrecon.nii.gz"
))
else:
fpet = os.path.join(
opth, fout + f'_itr{k+1}{fcomment}_inrecon.nii.gz')
nimpa.array2nii(im[::-1, ::-1, :], B, fpet)
log.info('recon time: %.3g', time.time() - stime)
# ========================================================================
log.info('applying decay correction of: %r', dcycrr)
log.info('applying quantification factor: %r to the whole image', qf)
log.info('for the frame duration of: %r', hst['dur'])
# additional factor for making it quantitative in absolute terms (derived from measurements)
img *= dcycrr * qf * qf_loc
# ---- save images -----
# -first convert to standard mMR image size
im = mmrimg.convert2e7(img, Cnt)
# -description text to NIfTI
# -attenuation number: if only bed present then it is 0.5
attnum = (1 * (np.sum(muh) > 0.5) + 1 * (np.sum(muo) > 0.5)) / 2.
descrip = (f"alg=osem"
f";sub=14"
f";att={attnum*(recmod>0)}"
f";sct={1*(recmod>1)}"
f";spn={Cnt['SPN']}"
f";itr={itr}"
f";fwhm=0"
f";t0={hst['t0']}"
f";t1={hst['t1']}"
f";dur={hst['dur']}"
f";qf={qf}")
# > file name of the output reconstructed image
# > (maybe used later even if not stored now)
if fout is None:
fpet = os.path.join(
opth,
(os.path.basename(datain['lm_bf']).split('.')[0] +
f"{frmno}_t{hst['t0']}-{hst['t1']}sec_itr{itr}{fcomment}.nii.gz"))
else:
fpet = os.path.join(opth, fout + f'_itr{itr}{fcomment}.nii.gz')
if store_img:
log.info('saving image to: %s', fpet)
nimpa.array2nii(im[::-1, ::-1, :], B, fpet, descrip=descrip)
im_smo = None
fsmo = None
if fwhm > 0:
im_smo = ndi.filters.gaussian_filter(im,
fwhm2sig(fwhm,
voxsize=Cnt['SZ_VOXY'] *
10),
mode='mirror')
if store_img:
fsmo = fpet.split('.nii.gz')[0] + '_smo-' + str(fwhm).replace(
'.', '-') + 'mm.nii.gz'
log.info('saving smoothed image to: ' + fsmo)
descrip.replace(';fwhm=0', ';fwhm=str(fwhm)')
nimpa.array2nii(im_smo[::-1, ::-1, :], B, fsmo, descrip=descrip)
# returning:
# (0) E7 image [can be smoothed];
# (1) file name of saved E7 image
# (2) [optional] scatter sino
# (3) [optional] single slice rebinned scatter
# (4) [optional] mask for scatter scaling based on attenuation data
# (5) [optional] random sino
# if ret_sinos and recmod>=3:
# recout = namedtuple('recout', 'im, fpet, ssn, sssr, amsk, rsn')
# recout.im = im
# recout.fpet = fout
# recout.ssn = ssn
# recout.sssr = sssr
# recout.amsk = amsk
# recout.rsn = rsino
# else:
# recout = namedtuple('recout', 'im, fpet')
# recout.im = im
# recout.fpet = fout
if ret_sinos and recmod >= 3 and itr > 1:
RecOut = namedtuple(
'RecOut', 'im, fpet, imsmo, fsmo, affine, ssn, sssr, amsk, rsn')
recout = RecOut(im, fpet, im_smo, fsmo, B, ssn, sct['ssrb'],
sct['mask'], rsino)
else:
RecOut = namedtuple('RecOut', 'im, fpet, imsmo, fsmo, affine')
recout = RecOut(im, fpet, im_smo, fsmo, B)
return recout
def simulate_recon(
measured_sino,
ctim,
scanner_params,
simulate_3d=False,
nitr=60,
fwhm_rm=0.,
slice_idx=-1,
randoms=None,
scatter=None,
mu_input=False,
msk_radius=29.,
psf=None,
):
'''
Reconstruct PET image from simulated input data
using the EM-ML (2D) or OSEM (3D) algorithm.
measured_sino : simulated emission data with photon attenuation
ctim : either a 2D CT image or a 3D CT image from which a 2D slice
is chosen (slice_idx) for estimation of the attenuation factors
slice_idx : index to extract one 2D slice for this simulation
if input image is 3D
nitr : number of iterations used for the EM-ML reconstruction algorithm
scanner_params : scanner parameters containing scanner constants and
axial and transaxial look up tables (LUTs)
randoms : randoms and scatter events (optional)
'''
# > decompose the scanner constants and LUTs for easier access
Cnt = scanner_params['Cnt']
txLUT = scanner_params['txLUT']
axLUT = scanner_params['axLUT']
psfkernel = mmrrec.psf_config(psf, Cnt)
if simulate_3d:
if ctim.ndim!=3 \
or ctim.shape!=(Cnt['SO_IMZ'], Cnt['SO_IMY'], Cnt['SO_IMX']):
raise ValueError(
'The CT/mu-map image does not match the scanner image shape.')
else:
# > 2D case with reduced rings
if len(ctim.shape) == 3:
# make sure that the shape of the input image matches the image size of the scanner
if ctim.shape[1:] != (Cnt['SO_IMY'], Cnt['SO_IMX']):
raise ValueError(
'The input image shape for x and y does not match the scanner image size.'
)
# pick the right slice index (slice_idx) if not given or mistaken
if slice_idx < 0:
log.warning(
'the axial index <slice_idx> is chosen to be in the middle of axial FOV.'
)
slice_idx = ctim.shape[0] / 2
if slice_idx >= ctim.shape[0]:
raise ValueError(
'The axial index for 2D slice selection is outside the image.'
)
elif len(ctim.shape) == 2:
# make sure that the shape of the input image matches the image size of the scanner
if ctim.shape != (Cnt['SO_IMY'], Cnt['SO_IMX']):
raise ValueError(
'The input image shape for x and y does not match the scanner image size.'
)
ctim.shape = (1, ) + ctim.shape
slice_idx = 0
if 'rSZ_IMZ' not in Cnt:
raise ValueError('Missing reduced axial FOV parameters.')
# --------------------
if mu_input:
mui = ctim
else:
# > get the mu-map [1/cm] from CT [HU]
mui = nimpa.ct2mu(ctim)
# > get rid of negative values
mui[mui < 0] = 0
# --------------------
if simulate_3d:
rmu = mui
# > number of axial sinograms
nsinos = Cnt['NSN11']
else:
# --------------------
# > create a number of slides of the same chosen image slice
# for reduced (fast) 3D simulation
rmu = mui[slice_idx, :, :]
rmu.shape = (1, ) + rmu.shape
rmu = np.repeat(rmu, Cnt['rSZ_IMZ'], axis=0)
# --------------------
# > number of axial sinograms
nsinos = Cnt['rNSN1']
# import pdb; pdb.set_trace()
# > attenuation factor sinogram
attsino = mmrprj.frwd_prj(rmu,
scanner_params,
attenuation=True,
dev_out=True)
nrmsino = np.ones(attsino.shape, dtype=np.float32)
# > randoms and scatter put together
if isinstance(randoms,
np.ndarray) and measured_sino.shape == randoms.shape:
rsng = mmraux.remgaps(randoms, txLUT, Cnt)
else:
rsng = 1e-5 * np.ones((Cnt['Naw'], nsinos), dtype=np.float32)
if isinstance(scatter,
np.ndarray) and measured_sino.shape == scatter.shape:
ssng = mmraux.remgaps(scatter, txLUT, Cnt)
else:
ssng = 1e-5 * np.ones((Cnt['Naw'], nsinos), dtype=np.float32)
# resolution modelling
Cnt['SIGMA_RM'] = mmrrec.fwhm2sig(fwhm_rm, voxsize=Cnt['SZ_VOXZ'] *
10) if fwhm_rm else 0
if simulate_3d:
log.debug('------ OSEM (%d) -------', nitr)
# measured sinogram in GPU-enabled shape
psng = mmraux.remgaps(measured_sino.astype(np.uint16), txLUT, Cnt)
# > mask for reconstructed image. anything outside it is set to zero
msk = mmrimg.get_cylinder(
Cnt, rad=msk_radius, xo=0, yo=0, unival=1, gpu_dim=True) > 0.9
# > init image
eimg = np.ones((Cnt['SZ_IMY'], Cnt['SZ_IMX'], Cnt['SZ_IMZ']),
dtype=np.float32)
# ------------------------------------
Sn = 14 # number of subsets
# -get one subset to get number of projection bins in a subset
Sprj, s = mmrrec.get_subsets14(0, scanner_params)
Nprj = len(Sprj)
# > init subset array and sensitivity image for a given subset
sinoTIdx = np.zeros((Sn, Nprj + 1), dtype=np.int32)
# > init sensitivity images for each subset
sim = np.zeros((Sn, Cnt['SZ_IMY'], Cnt['SZ_IMX'], Cnt['SZ_IMZ']),
dtype=np.float32)
tmpsim = cu.zeros((Cnt['SZ_IMY'], Cnt['SZ_IMX'], Cnt['SZ_IMZ']),
dtype=np.float32)
for n in trange(Sn,
desc="sensitivity",
leave=log.getEffectiveLevel() < logging.INFO):
# first number of projection for the given subset
sinoTIdx[n, 0] = Nprj
sinoTIdx[n, 1:], s = mmrrec.get_subsets14(n, scanner_params)
# > sensitivity image
petprj.bprj(tmpsim.cuvec,
cu.asarray(attsino[sinoTIdx[n, 1:], :]).cuvec, txLUT,
axLUT, sinoTIdx[n, 1:], Cnt)
sim[n] = tmpsim
del tmpsim
# -------------------------------------
for _ in trange(nitr,
desc="OSEM",
disable=log.getEffectiveLevel() > logging.INFO,
leave=log.getEffectiveLevel() < logging.INFO):
petprj.osem(eimg, psng, rsng, ssng, nrmsino, attsino, sinoTIdx,
sim, msk, psfkernel, txLUT, axLUT, Cnt)
eim = mmrimg.convert2e7(eimg, Cnt)
else:
def psf(x, output=None):
if Cnt['SIGMA_RM']:
x = ndi.gaussian_filter(x,
sigma=Cnt['SIGMA_RM'],
mode='constant',
output=None)
return x
# > estimated image, initialised to ones
eim = np.ones(rmu.shape, dtype=np.float32)
msk = mmrimg.get_cylinder(
Cnt, rad=msk_radius, xo=0, yo=0, unival=1, gpu_dim=False) > 0.9
# > sensitivity image for the EM-ML reconstruction
sim = mmrprj.back_prj(attsino, scanner_params)
sim_inv = 1 / psf(sim)
sim_inv[~msk] = 0
rndsct = rsng + ssng
for _ in trange(nitr,
desc="MLEM",
disable=log.getEffectiveLevel() > logging.INFO,
leave=log.getEffectiveLevel() < logging.INFO):
# > remove gaps from the measured sinogram
# > then forward project the estimated image
# > after which divide the measured sinogram
# by the estimated sinogram (forward projected)
crrsino = (
mmraux.remgaps(measured_sino, txLUT, Cnt) /
(mmrprj.frwd_prj(psf(eim), scanner_params, dev_out=True) +
rndsct))
# > back project the correction factors sinogram
bim = mmrprj.back_prj(crrsino, scanner_params)
bim = psf(bim, output=bim)
# > divide the back-projected image by the sensitivity image
# > update the estimated image and remove NaNs
eim *= bim * sim_inv
eim[np.isnan(eim)] = 0
return eim
def vsm(
datain,
mumaps,
em,
scanner_params,
histo=None,
rsino=None,
prcnt_scl=0.1,
fwhm_input=0.42,
mask_threshlod=0.999,
snmsk=None,
emmsk=False,
interpolate=True,
return_uninterp=False,
return_ssrb=False,
return_mask=False,
return_scaling=False,
scaling=True,
self_scaling=False,
save_sax=False,
):
'''
Voxel-driven scatter modelling (VSM).
Obtain a scatter sinogram using the mu-maps (hardware and object mu-maps)
an estimate of emission image, the prompt measured sinogram, an
estimate of the randoms sinogram and a normalisation sinogram.
Input:
- datain: Contains the data used for scatter-specific detector
normalisation. May also include the non-corrected
emission image used for masking, when requested.
- mumaps: A tuple of hardware and object mu-maps (in this order).
- em: An estimate of the emission image.
- histo: Dictionary containing the histogrammed measured data into
sinograms.
- rsino: Randoms sinogram (3D). Needed for proper scaling of
scatter to the prompt data.
- scanner_params: Scanner specific parameters.
- prcnt_scl: Ratio of the maximum scatter intensities below which the
scatter is not used for fitting it to the tails of prompt
data. Default is 10%.
- emmsk: When 'True' it will use uncorrected emission image for
masking the sources (voxels) of photons to be used in the
scatter modelling.
- scaling: performs scaling to the data (sinogram)
- self_scaling: Scaling is performed on span-1 without the help of SSR
scaling and using the sax factors (scatter axial factors).
If False (default), the sax factors have to be provided.
- sax: Scatter axial factors used for scaling with SSR sinograms.
'''
# > decompose constants, transaxial and axial LUTs are extracted
Cnt = scanner_params['Cnt']
txLUT = scanner_params['txLUT']
axLUT = scanner_params['axLUT']
if self_scaling:
scaling = True
# > decompose mu-maps
muh, muo = mumaps
if emmsk and not os.path.isfile(datain['em_nocrr']):
log.info(
'reconstructing emission data without scatter and attenuation corrections'
' for mask generation...')
recnac = mmrrec.osemone(datain,
mumaps,
histo,
scanner_params,
recmod=0,
itr=3,
fwhm=2.0,
store_img=True)
datain['em_nocrr'] = recnac.fpet
# if rsino is None and not histo is None and 'rsino' in histo:
# rsino = histo['rsino']
# > if histogram data or randoms sinogram not given, then no scaling or normalisation
if (histo is None) or (rsino is None):
scaling = False
# -get the normalisation components
nrmcmp, nhdr = mmrnorm.get_components(datain, Cnt)
# -smooth for defining the sino scatter only regions
if fwhm_input > 0.:
mu_sctonly = ndi.filters.gaussian_filter(mmrimg.convert2dev(muo, Cnt),
fwhm2sig(fwhm_input, Cnt),
mode='mirror')
else:
mu_sctonly = muo
if Cnt['SPN'] == 1:
snno = Cnt['NSN1']
snno_ = Cnt['NSN64']
ssrlut = axLUT['sn1_ssrb']
saxnrm = nrmcmp['sax_f1']
elif Cnt['SPN'] == 11:
snno = Cnt['NSN11']
snno_ = snno
ssrlut = axLUT['sn11_ssrb']
saxnrm = nrmcmp['sax_f11']
# LUTs for scatter
sctLUT = get_sctLUT(scanner_params)
# > smooth before scaling/down-sampling the mu-map and emission images
if fwhm_input > 0.:
muim = ndi.filters.gaussian_filter(muo + muh,
fwhm2sig(fwhm_input, Cnt),
mode='mirror')
emim = ndi.filters.gaussian_filter(em,
fwhm2sig(fwhm_input, Cnt),
mode='mirror')
else:
muim = muo + muh
emim = em
muim = ndi.interpolation.zoom(muim, Cnt['SCTSCLMU'],
order=3) # (0.499, 0.5, 0.5)
emim = ndi.interpolation.zoom(emim, Cnt['SCTSCLEM'],
order=3) # (0.34, 0.33, 0.33)
# -smooth the mu-map for mask creation.
# the mask contains voxels for which attenuation ray LUT is found.
if fwhm_input > 0.:
smomu = ndi.filters.gaussian_filter(muim,
fwhm2sig(fwhm_input, Cnt),
mode='mirror')
mumsk = np.int8(smomu > 0.003)
else:
mumsk = np.int8(muim > 0.001)
# CORE SCATTER ESTIMATION
NSCRS, NSRNG = sctLUT['NSCRS'], sctLUT['NSRNG']
sctout = {
'sct_3d':
np.zeros((Cnt['TOFBINN'], snno_, NSCRS, NSCRS), dtype=np.float32),
'sct_val':
np.zeros((Cnt['TOFBINN'], NSRNG, NSCRS, NSRNG, NSCRS),
dtype=np.float32)
}
# <<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>>
nifty_scatter.vsm(sctout, muim, mumsk, emim, sctLUT, axLUT, Cnt)
# <<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>>
sct3d = sctout['sct_3d']
sctind = sctLUT['sct2aw']
log.debug('total scatter sum: {}'.format(np.sum(sct3d)))
# -------------------------------------------------------------------
# > initialise output dictionary
out = {}
if return_uninterp:
out['uninterp'] = sct3d
out['indexes'] = sctind
# -------------------------------------------------------------------
if np.sum(sct3d) < 1e-04:
log.warning('total scatter below threshold: {}'.format(np.sum(sct3d)))
sss = np.zeros((snno, Cnt['NSANGLES'], Cnt['NSBINS']),
dtype=np.float32)
asnmsk = np.zeros((snno, Cnt['NSANGLES'], Cnt['NSBINS']),
dtype=np.float32)
sssr = np.zeros((Cnt['NSEG0'], Cnt['NSANGLES'], Cnt['NSBINS']),
dtype=np.float32)
return sss, sssr, asnmsk
# import pdb; pdb.set_trace()
# -------------------------------------------------------------------
if interpolate:
# > interpolate basic scatter distributions into full size and
# > transfer them to sinograms
log.debug('transaxial scatter interpolation...')
start = time.time()
ssn, sssr = intrp_bsct(sct3d, Cnt, sctLUT, ssrlut)
stop = time.time()
log.debug('scatter interpolation done in {} sec.'.format(stop - start))
if not scaling:
out['ssrb'] = sssr
out['sino'] = ssn
return out
else:
return out
# -------------------------------------------------------------------
# -------------------------------------------------------------------
# import pdb; pdb.set_trace()
'''
debugging scatter:
import matplotlib.pyplot as plt
ss = np.squeeze(sct3d)
ss = np.sum(ss, axis=0)
plt.matshow(ss)
plt.matshow(sct3d[0,41,...])
plt.matshow(np.sum(sct3d[0,0:72,...],axis=0))
plt.plot(np.sum(sct3d, axis=(0,2,3)))
rslt = sctout['sct_val']
rslt.shape
plt.matshow(rslt[0,4,:,4,:])
debugging scatter:
plt.matshow(np.sum(sssr, axis=(0,1)))
plt.matshow(np.sum(ssn, axis=(0,1)))
plt.matshow(sssr[0,70,...])
plt.matshow(sssr[0,50,...])
'''
# -------------------------------------------------------------------
# > get SSR for randoms from span-1 or span-11
rssr = np.zeros((Cnt['NSEG0'], Cnt['NSANGLES'], Cnt['NSBINS']),
dtype=np.float32)
if scaling:
for i in range(snno):
rssr[ssrlut[i], :, :] += rsino[i, :, :]
# ATTENUATION FRACTIONS for scatter only regions, and NORMALISATION for all SCATTER
# <<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>>
currentspan = Cnt['SPN']
Cnt['SPN'] = 1
atto = cu.zeros((txLUT['Naw'], Cnt['NSN1']), dtype=np.float32)
petprj.fprj(atto.cuvec,
cu.asarray(mu_sctonly).cuvec, txLUT, axLUT,
np.array([-1], dtype=np.int32), Cnt, 1)
atto = mmraux.putgaps(atto, txLUT, Cnt)
# --------------------------------------------------------------
# > get norm components setting the geometry and axial to ones
# as they are accounted for differently
nrmcmp['geo'][:] = 1
nrmcmp['axe1'][:] = 1
# get sino with no gaps
nrmg = np.zeros((txLUT['Naw'], Cnt['NSN1']), dtype=np.float32)
mmr_auxe.norm(nrmg, nrmcmp, histo['buckets'], axLUT, txLUT['aw2ali'], Cnt)
nrm = mmraux.putgaps(nrmg, txLUT, Cnt)
# --------------------------------------------------------------
# > get attenuation + norm in (span-11) and SSR
attossr = np.zeros((Cnt['NSEG0'], Cnt['NSANGLES'], Cnt['NSBINS']),
dtype=np.float32)
nrmsssr = np.zeros((Cnt['NSEG0'], Cnt['NSANGLES'], Cnt['NSBINS']),
dtype=np.float32)
for i in range(Cnt['NSN1']):
si = axLUT['sn1_ssrb'][i]
attossr[si, :, :] += atto[i, :, :] / float(axLUT['sn1_ssrno'][si])
nrmsssr[si, :, :] += nrm[i, :, :] / float(axLUT['sn1_ssrno'][si])
if currentspan == 11:
Cnt['SPN'] = 11
nrmg = np.zeros((txLUT['Naw'], snno), dtype=np.float32)
mmr_auxe.norm(nrmg, nrmcmp, histo['buckets'], axLUT, txLUT['aw2ali'],
Cnt)
nrm = mmraux.putgaps(nrmg, txLUT, Cnt)
# --------------------------------------------------------------
# <<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>>
# get the mask for the object from uncorrected emission image
if emmsk and os.path.isfile(datain['em_nocrr']):
nim = nib.load(datain['em_nocrr'])
eim = nim.get_fdata(dtype=np.float32)
eim = eim[:, ::-1, ::-1]
eim = np.transpose(eim, (2, 1, 0))
em_sctonly = ndi.filters.gaussian_filter(eim,
fwhm2sig(.6, Cnt),
mode='mirror')
msk = np.float32(em_sctonly > 0.07 * np.max(em_sctonly))
msk = ndi.filters.gaussian_filter(msk,
fwhm2sig(.6, Cnt),
mode='mirror')
msk = np.float32(msk > 0.01)
msksn = mmrprj.frwd_prj(msk, txLUT, axLUT, Cnt)
mssr = mmraux.sino2ssr(msksn, axLUT, Cnt)
mssr = mssr > 0
else:
mssr = np.zeros((Cnt['NSEG0'], Cnt['NSANGLES'], Cnt['NSBINS']),
dtype=bool)
# <<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>><<+>>
# ======= SCALING ========
# > scale scatter using non-TOF SSRB sinograms
# > gap mask
rmsk = (txLUT['msino'] > 0).T
rmsk.shape = (1, Cnt['NSANGLES'], Cnt['NSBINS'])
rmsk = np.repeat(rmsk, Cnt['NSEG0'], axis=0)
# > include attenuating object into the mask (and the emission if selected)
amsksn = np.logical_and(attossr >= mask_threshlod, rmsk) * ~mssr
# > scaling factors for SSRB scatter
scl_ssr = np.zeros((Cnt['NSEG0']), dtype=np.float32)
for sni in range(Cnt['NSEG0']):
# > region for scaling defined by the percentage of lowest
# > but usable/significant scatter
thrshld = prcnt_scl * np.max(sssr[sni, :, :])
amsksn[sni, :, :] *= (sssr[sni, :, :] > thrshld)
amsk = amsksn[sni, :, :]
# > normalised estimated scatter
mssn = sssr[sni, :, :] * nrmsssr[sni, :, :]
vpsn = histo['pssr'][sni, amsk] - rssr[sni, amsk]
scl_ssr[sni] = np.sum(vpsn) / np.sum(mssn[amsk])
# > scatter SSRB sinogram output
sssr[sni, :, :] *= nrmsssr[sni, :, :] * scl_ssr[sni]
# === scale scatter for the full-size sinogram ===
sss = np.zeros((snno, Cnt['NSANGLES'], Cnt['NSBINS']), dtype=np.float32)
for i in range(snno):
sss[i, :, :] = ssn[i, :, :] * scl_ssr[ssrlut[i]] * saxnrm[i] * nrm[
i, :, :]
'''
# > debug
si = 60
ai = 60
matshow(sssr[si,...])
figure()
plot(histo['pssr'][si,ai,:])
plot(rssr[si,ai,:]+sssr[si,ai,:])
plot(np.sum(histo['pssr'],axis=(0,1)))
plot(np.sum(rssr+sssr,axis=(0,1)))
'''
# === OUTPUT ===
if return_uninterp:
out['uninterp'] = sct3d
out['indexes'] = sctind
if return_ssrb:
out['ssrb'] = sssr
out['rssr'] = rssr
if return_mask:
out['mask'] = amsksn
if return_scaling:
out['scaling'] = scl_ssr
# if self_scaling:
# out['scl_sn1'] = scl_ssn
if not out:
return sss
else:
out['sino'] = sss
return out
