Correcting fast irregular motion in PET: maximum-likelihood motion and activity (MLMA) reconstruction

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Reconstruction

Background:

• Positron emission tomography (PET) imaging is susceptible to motion blur due to lengthy acquisition times.
• Traditional motion correction methods struggle with irregular motion and high noise in short timeframes.
• Compensating for motion is crucial for enhancing PET image quality.

Purpose of the Study:

• To develop a novel method for motion compensation in PET imaging.
• To address limitations of traditional motion correction techniques, particularly for irregular and high-frequency motion.
• To introduce Maximum-Likelihood Motion and Activity (MLMA) reconstruction for improved PET image quality.

Main Methods:

• Proposed a new alternating estimation and correction method for high-temporal-frequency motion in PET.
• Utilized a cubic B-spline motion model and spatial regularization for motion compressibility and smoothness.
• Configured MLMA at 2 Hz resolution and validated on digital phantoms, anthropomorphic torso phantom, and clinical patient data.

Main Results:

• MLMA accurately corrected high-frequency motion (2 Hz) with subvoxel accuracy (up to 2.5 mm RMSE).
• The method captured realistic breathing motion (14.7 mm amplitude, 4.5 s period).
• Visually significant improvements in PET image quality were observed.

Conclusions:

• The MLMA reconstruction method effectively resolves motion in short PET timeframes.
• It overcomes challenges posed by low counts and inherent noise in PET projection data.
• MLMA enables visualization of anatomical structures with temporal motion information.