def scan_and_reconstruct(photons, material, phantom, scale, angles, mas=10000, alpha=0.001, correct=True): """ Simulation of the CT scanning process reconstruction = scan_and_reconstruct(photons, material, phantom, scale, angles, mas, alpha) takes the phantom data in phantom (samples x samples), scans it using the source photons and material information given, as well as the scale (in cm), number of angles, time-current product in mas, and raised-cosine power alpha for filtering. The output reconstruction is the same size as phantom.""" # convert source (photons per (mas, cm^2)) to photons # create sinogram from phantom data, with received detector values sinogram = ct_scan(photons, material, phantom, scale, angles, mas) # convert detector values into calibrated attenuation values sinogram = ct_calibrate(photons, material, sinogram, scale, correct) # Ram-Lak sinogram = ramp_filter(sinogram, scale, alpha) # Back-projection reconstruction = back_project(sinogram) # convert to Hounsfield Units reconstruction = hu(photons, material, reconstruction, scale) return reconstruction
def scan_and_backproject(photon_source,
material_data,
phantom,
scale,
angles,
mas=10000,
alpha=0.001):
"""
Scan and backproject without convolving with a RamLak filter. Use for report and demonstration only.
"""
# create sinogram from phantom data, with received detector values
sinogram = ct_scan(photon_source,
material_data,
phantom,
scale,
angles,
interpolation_order=1)
# convert detector values into calibrated attenuation values
total_attenuation = ct_calibrate(photon_source, material_data, sinogram,
scale)
# Back-projection
backprojection = back_project(total_attenuation)
return backprojection
def reconstruct_all(self, file, method=None, alpha=None):
""" reconstruct_all( FILENAME, ALPHA ) creates a series of DICOM
files for the Xtreme RSQ data. FILENAME is the base file name for
the data, and ALPHA is the power of the raised cosine function
used to filter the data.
reconstruct_all( FILENAME, ALPHA, METHOD ) can be used to
specify how the data is reconstructed. Possible options are:
'parallel' - reconstruct each slice separately using a fan to parallel
conversion
'fdk' - approximate FDK algorithm for better reconstruction"""
if alpha is None:
alpha = 0.001
if method is None:
method = 'parallel'
# set frame number and DICOM UIDs for saving to multiple frames
z = 1
seriesuid = pydicom.uid.generate_uid()
studyuid = pydicom.uid.generate_uid()
frameuid = pydicom.uid.generate_uid()
time = datetime.datetime.now()
# main loop over each z-fan
for fan in range(0, self.scans, self.fan_scans):
if method == 'fdk':
# correct reconstruction using FDK method, self.fan_scans scans at a time
pass
else:
# default method should reconstruct each slice separately
sys.stdout.write('\n\n\n\n\n')
for scan in range(fan + self.skip_scans,
fan + self.fan_scans - self.skip_scans):
if (scan < self.scans):
sys.stdout.write(
'\x1b[2K\x1b[1A\x1b[2K\x1b[1A\x1b[2K\x1b[1A\x1b[2K\x1b[1A\x1b[1A'
+ "Fan: " + str(fan // self.fan_scans + 1) + '\n')
sys.stdout.write("Scan: " + str(
(scan - self.skip_scans) % self.fan_scans + 1) +
'\n')
# get scan detector values, noise floor and reference
sinogram, noise, ref = self.get_rsq_slice(scan)
# convert detector values into calibrated attenuation values
sinogram = -np.log((sinogram - noise) / (ref - noise))
# convert scan from fan to parallel
sinogram = self.fan_to_parallel(sinogram)
# apply Ram-Lak filter
sinogram = ramp_filter(sinogram, self.scale, alpha)
# back project
reconstruction = back_project(sinogram)
# convert to Hounsfield units
reconstruction = 1000 * (reconstruction -
23.835e-3) / 23.835e-3
reconstruction = reconstruction.clip(min=-1024,
max=3071)
# save as dicom file
create_dicom(reconstruction, file, self.scale,
self.scale, z, studyuid, seriesuid,
frameuid, time)
z = z + 1
return
def reconstruct_all(self, file, method='parallel', alpha=0.001):
""" reconstruct_all( FILENAME, ALPHA ) creates a series of DICOM
files for the Xtreme RSQ data. FILENAME is the base file name for
the data, and ALPHA is the power of the raised cosine function
used to filter the data."""
# set frame number and DICOM UIDs for saving to multiple frames
z = 1
seriesuid = pydicom.uid.generate_uid()
studyuid = pydicom.uid.generate_uid()
frameuid = pydicom.uid.generate_uid()
time = datetime.datetime.now()
# main loop over each z-fan
for fan in range(0, self.scans, self.fan_scans):
# reconstruct each slice separately
for scan in range(fan + self.skip_scans,
fan + self.fan_scans - self.skip_scans):
if (scan < self.scans):
# obtain slice data
f, fmin, fmax = self.get_rsq_slice(scan)
# remove noise
f = f - fmin
fmax = fmax - fmin
# calibrate
f = -np.log(f / fmax)
# convert to a parallel sinogram
p = self.fan_to_parallel(f)
# filter sinogram (FBP)
scale = self.scale / 10
filtered = ramp_filter(p, scale)
# reconstruct
reconstruction = back_project(filtered)
# convert to Hu
reconstruction = (reconstruction - 0.2192) * 1000 / 0.2192
reconstruction = np.where(reconstruction < -1024, -1024,
reconstruction)
reconstruction = np.where(reconstruction > 3072, 3072,
reconstruction)
plt.close()
draw(reconstruction)
# save as dicom file
"""create_dicom(x, filename, sp, sz, f) creates a new DICOM file with a
name formed from the given filename and the frame number f, and
containing data from x. The pixel scale is given by sp, and the frame
spacing is given by sz, both of which are in mm."""
create_dicom(reconstruction,
file,
sp=scale,
sz=scale,
f=z,
storage_directory='results',
study_uid=studyuid,
series_uid=seriesuid,
time=time)
z = z + 1
return