An objective evaluation method for head motion estimation in PET-Motion corrected centroid-of-distribution

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Neuroscience

Background:

• Head motion significantly impacts brain PET study quality.
• Current motion correction (MC) evaluation methods lack objectivity due to confounding factors and reliance on final image metrics like standardized uptake value (SUV).
• Objective quality control for motion information is crucial in real-world PET studies.

Purpose of the Study:

• To introduce a novel algorithm, motion corrected centroid-of-distribution (MCCOD), for objective quality control of rigid motion information in brain PET.
• To enable evaluation of MC effectiveness using only raw PET data, independent of image reconstruction.
• To validate MCCOD's ability to detect motion tracking errors.

Main Methods:

• Developed MCCOD, a 3D trace of tracer distribution center after rigid MC, using only PET raw data.
• Implemented a deep learning-based brain mask segmentation for non-time-of-flight (TOF) MCCOD generation.
• Validated MCCOD using simulations and tested on real TOF and non-TOF PET data.

Main Results:

• MCCOD effectively detects abrupt translation motion errors, even with slowly varying tracer distributions.
• Inaccurate motion information leads to abrupt changes in the MCCOD trace, indicating poor motion correction quality.
• MCCOD proved effective on both TOF and non-TOF scanners, with deep learning segmentation being essential for non-TOF data.

Conclusions:

• MCCOD provides an objective, raw-data-based method for quality control of motion information in brain PET.
• This algorithm can be used to compare different motion estimation methods and enhance existing motion information.
• MCCOD addresses a critical need for reliable MC evaluation in clinical brain PET studies.