Improved frame-based estimation of head motion in PET brain imaging.

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Neuroscience

Background:

• Head motion during Positron Emission Tomography (PET) brain imaging significantly degrades image quality and quantitation.
• Existing head restraints are unreliable, necessitating advanced motion compensation strategies.
• Data-driven motion estimation and external tracking are explored as alternatives.

Purpose of the Study:

• To introduce a novel data-driven motion estimation method for PET brain imaging.
• To reduce image quality degradation caused by intraframe motion.
• To improve quantitation in PET brain studies through motion compensation.

Main Methods:

• PET list mode data divided into 5-second frames, reconstructed without attenuation correction.
• 3D multiresolution registration algorithm used for interframe motion estimation and compensation.
• Simulated head motion introduced to PET data to validate the method.

Main Results:

• The method accurately compensates for gradual and step-like motions with 5-second frames (0.2 mm average spatial accuracy).
• Complex six-degree-of-freedom motion estimated with 0.3 mm average accuracy.
• Preprocessing of 5-second images is crucial for successful registration; method is robust to CT-PET timing variations.

Conclusions:

• Motion can be effectively estimated for short 5-second frames in FDG PET brain imaging.
• Utilizing non-attenuation corrected frames enhances robustness against motion-induced errors.
• Longer frame times (60 seconds) lead to significant accuracy degradation (approx. 2 mm).