Evaluating different methods of MR-based motion correction in simultaneous PET/MR using a head phantom moved by a...

Area of Science:

• Medical Imaging
• Neuroscience
• Biophysics

Background:

• Patient motion during simultaneous Positron Emission Tomography and Magnetic Resonance (PET/MR) imaging leads to artifacts, compromising image quality and diagnostic accuracy.
• Simultaneous PET/MR systems offer MR-based motion registration for PET data correction.
• Reproducible phantom-based motion is crucial for evaluating motion correction effectiveness.

Purpose of the Study:

• To establish a phantom-based experimental setup for reproducible motion generation in PET/MR imaging of the human head.
• To evaluate and compare different motion correction strategies using both vendor-provided and open-source software.
• To assess the effectiveness of motion correction on PET data validity.

Main Methods:

• An MR-compatible robotic system generated rigid movements of a head phantom.
• Motion was estimated and corrected using various tools, including SIRF with SPM, NiftyReg, MCFLIRT, and BrainCompass, leveraging both PET and simultaneous MR data.
• Correction effectiveness was quantified by analyzing the segmented volume, center-of-mass, and maximum activity concentration of an internal phantom component.

Main Results:

• Robot-induced motion caused volume increases ranging from 2.7% to 36.3%.
• Motion correction using BrainCompass reduced volume increase to 0.7-4.7%, while MCFLIRT achieved -2.8-0.4%.
• MCFLIRT demonstrated superior performance in center-of-mass accuracy (0.2-0.6 mm deviation) compared to BrainCompass (0.5-1.8 mm).

Conclusions:

• The developed experimental setup reliably generates reproducible motion patterns for PET/MR research.
• Open-source motion correction software provides a viable and effective alternative to vendor-specific solutions.
• This work validates a method for assessing motion correction efficacy in head PET/MR imaging.