Gating, enhanced gating, and beyond: information utilization strategies for motion management, applied to preclinical
Adam Leon Kesner1, Galith Abourbeh, Eyal Mishani
1Department of Medical Biophysics and Nuclear Medicine, Kiryat Hadassah, P,O, Box 12000, Jerusalem 91120, Israel. adam.kesner@fulbrightmail.org.
Background:
Respiratory gating and gate optimization strategies present solutions for overcoming image degradation caused by respiratory motion in PET and traditionally utilize hardware systems and/or employ complex processing algorithms. In this work, we aimed to advance recently emerging data-driven gating methods and introduce a new strategy for optimizing the four-dimensional data based on information contained in that data. These algorithms are combined to form an automated motion correction workflow.
Methods:
Software-based gating methods were applied to a nonspecific population of 84 small-animal rat PET scans to create respiratory gated images. The gated PET images were then optimized using an algorithm we introduce as 'gating+' to reduce noise and optimize signal; the technique was also tested using simulations. Gating+ is based on a principle of only using gated information if and where it adds a net benefit, as evaluated in temporal frequency space. Motion-corrected images were assessed quantitatively and qualitatively.
Results:
Of the small-animal PET scans, 71% exhibited quantifiable motion after software gating. The mean liver displacement was 3.25 mm for gated and 3.04 mm for gating+ images. The (relative) mean percent standard deviations measured in background ROIs were 1.53, 1.05, and 1.00 for the gated, gating+, and ungated values, respectively. Simulations confirmed that gating+ image voxels had a higher probability of being accurate relative to the corresponding ungated values under varying noise and motion scenarios. Additionally, we found motion mapping and phase decoupling models that readily extend from gating+ processing.
Conclusions:
Raw PET data contain information about motion that is not currently utilized. In our work, we showed that through automated processing of standard (ungated) PET acquisitions, (motion-) information-rich images can be constructed with minimal risk of noise introduction. Such methods have the potential for implementation with current PET technology in a robust and reproducible way.
