Motion compensation for PET-MR applications Christoph Kolbitsch
Motion compensation for
PET-MR applications
Christoph Kolbitsch
Overview
Gadget
Gadget
Gadget
Gadget
Gadget
Gadget
Gadget
Gadget
• Read data
• Remove oversampling in readout
• Apply k-space filter
• Transform k-space data to image space
• Read from raw matrix
• Conversion to uint
• Save data to hdf5
XML
Overview - filterRecon
• Read data
• Remove oversampling in readout
• Convert to uint
• Save data to hdf5
• Apply k-space filter
• Transform k-space data to image space
Gadgetron
Matlab
XML
filterRecon.m
filterRecon
Matlab
• Read data
• Remove oversampling in readout
• Apply k-space filter
• Transform from k-space to image space
Gadgetron
Matlab
• Define data and parameter files
• Receive image data
• Postprocessing
Matlab
Hypothetical PET-MR Compensation for MR motion
convert toISMRMRDk-space++
gadget 1
gadget
STIR
raw
ISMRMRD image or DICOM
[Image: Hartford Hospital]
other dataList mode / sinograms
gadget N
PET derived motion signal
DEMO: Applying Motion to MRI
gadget 1
gadget
ISMRMRD image or DICOM
ISMRMRDinspector
MATLABcode
gadget N
Modify recon to apply motion “correction” phases to k-space.(15mm translation, half-way through acquisition)
𝒆−𝟐𝝅𝒊𝒌∆𝒙
ISMRMRD file: simple_gre.h5Config: matlab_moco.xml
explore_h5.m
Iterative reconstruction
• Read data
• Remove oversampling in readout
• Reconstruct using GPU SENSE
• Convert and save data to hdf5
• Calculate k-space trajectory, dcf, csm,
regularisation,....
Gadgetron
Matlab
Iterative reconstruction
Iterative reconstruction
Gadgetron
Matlab
MatlabBufferGadget.cpp
Iterative reconstruction
Matlab
Gadgetron
MatlabBufferGadget.cpp
Iterative reconstruction
gpuRadialPrepGadget.cpp
Prep
SENSE
SPIRIT
Python
Python Notebook
‘Live’ code, run in cells.
Similar to running cells in MATLAB.
Jupyter Notebook