Frame-to-frame image realignment assessment tool for dynamic brain positron emission tomography.

Area of Science:

• Neuroimaging
• Medical Physics
• Radiochemistry

Background:

• Subject motion during Positron Emission Tomography (PET) brain scans degrades image quality.
• Motion artifacts can lead to inaccurate biological outcome measures, particularly with high-resolution data.
• Accurate frame-to-frame realignment is crucial for dynamic brain PET studies.

Purpose of the Study:

• To propose and evaluate a semiautomatic method for assessing the quality of frame-to-frame image realignments in dynamic brain PET.
• To ensure accurate compensation for subject motion during PET scans.
• To improve the reliability of biological outcome measures derived from brain PET data.

Main Methods:

• Developed a quality assessment method based on comparing two independent motion detection approaches (optical tracking vs. automated image registration).
• Calculated transformation matrices for frame-to-frame realignment using both Polaris Vicra and AIR algorithms.
• Defined a metric based on the agreement between transformation matrices, validated against visual inspection, to determine realignment accuracy.

Main Results:

• The proposed method successfully categorized 53% of image realignments as accurate.
• An 11% miscategorization rate was observed (6% of the total dataset).
• The assessment tool provided a 45% time saving compared to manual visual inspection.

Conclusions:

• The semiautomatic tool efficiently evaluates realignment success with comparable accuracy to manual inspection.
• It reduces operator time and can be implemented in centers with or without motion monitoring technology.
• The method is flexible for different tracers and analysis goals in dynamic brain PET.