Dynamic cardiac PET motion correction using 3D normalized gradient fields in patients and phantom simulations.

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Processing

Background:

• Motion artifacts, including whole-body motion and cardiac creep, significantly impact the accuracy of dynamic [N-13]-ammonia PET/CT studies.
• Accurate quantification of myocardial flow reserve is essential for diagnosing and managing cardiovascular diseases.

Purpose of the Study:

• To implement and evaluate a three-dimensional (3D) normalized gradient field algorithm for correcting motion and cardiac creep in [N-13]-ammonia PET/CT.
• To assess the accuracy of this motion correction technique using a dynamic phantom simulation model.

Main Methods:

• Developed a rigid-body algorithm using 3D normalized gradient fields with multi-resolution and off-voxel sampling to minimize artifacts.
• Employed a weighted similarity metric for optimization, avoiding segmentation and accounting for opposing gradients.
• Validated the algorithm on 43 retrospective [N-13]-ammonia PET/CT studies and a dynamic XCAT simulation.

Main Results:

• Observed a mean cranial-to-caudal translation of 7 mm during adenosine stress in patient studies.
• Demonstrated high correlation between noninvasive and invasive measures of relative and myocardial flow reserve.
• XCAT simulations showed motion correction accuracy with errors <1 mm in late frames and <3 mm in early frames.

Conclusions:

• [N-13]-ammonia PET/CT patients exhibit significant cranial-to-caudal translation due to cardiac creep, particularly during stress.
• Optimizing 3D normalized gradient fields accurately corrects this motion, improving diagnostic reliability.
• The proposed method is robust, handles opposing gradients without segmentation, and is effective despite PET-CT mismatch.