Quantitative validation of data-driven motion correction for brain PET using phantom with motion generator system.

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Positron Emission Tomography (PET)

Background:

• Head motion during brain PET scans degrades image quality and quantitative accuracy.
• A data-driven motion correction (MC) method using ultrafast list-mode reconstruction has shown promise.
• Evaluating MC with known motions and tracer distributions is crucial for clinical validation.

Purpose of the Study:

• To quantitatively evaluate a data-driven MC method for brain PET.
• To assess MC performance using a brain phantom with known tracer distributions.
• To validate MC accuracy with a motion generator system for known, variable motions.

Main Methods:

• A Hoffman 3D brain phantom was used with [18F]fluoro-2-deoxy-D-glucose (FDG) to simulate high (20 MBq) and low (3 MBq) radioactivity conditions.
• A custom motion generator simulated rotations (z-axis: 15°, 30°; x-axis: 3°, 6°) and translations (z-axis: 5 mm, 10 mm).
• Data-driven MC was applied with frame durations of 1, 2, 5, 10, and 20 s; estimated motions were compared to optical tracker data; %contrast and gray matter CV% were calculated.

Main Results:

• Motion estimation was stable under high-radioactivity but less stable under low-radioactivity conditions, especially with shorter frame durations.
• Longer frame durations led to underestimation of continuous motion.
• Data-driven MC improved %contrast and gray matter CV% across all conditions; shorter frames better corrected rapid motions but increased noise in low-radioactivity scans.

Conclusions:

• The data-driven MC method effectively enhances motion-affected brain PET image quality in both high- and low-radioactivity settings.
• The method demonstrates broad applicability for improving PET image analysis.
• Correction accuracy was found to be reduced under lower-radioactivity conditions.