Motion compensation for PET in TOF-PET/CT and TOF-PET/MR brain imaging

In PET/CT or PET/MR brain imaging, images of the metabolism and anatomy of the patient are acquired. If the patient moves during the PET acquisition, then motion blurring is caused, the image quality is deteriorated, and an accurate clinical diagnosis might become impossible. Several research groups (including our own) have previously developed marker based methods to track the motion of the head, such that the motion can be compensated for when the image is reconstructed from the raw data. This enables high quality imaging in subjects that are likely to move during the scan, such as children and people suffering from neurodegenerative or psychiatric disorders. However, this approach requires that markers are rigidly attached to the head of the patient. This is inconvenient and may not be well tolerated by all patients. State of the art PET systems have a time-of-flight (TOF) resolution of 400 picoseconds or less. In a preliminary study, we’ve found that this excellent TOF-resolution enables the determination of patient motion from the PET-data themselves, eliminating the need for markers and additional motion tracking hardware. In this project, we will use this approach to develop a clinical data processing procedure which produces motion-free PET images in patients who move their head during the PET scan. This procedure will be entirely based on image processing and has therefore no impact on the standard clinical scanning protocols. The new method will be validated in clinical PET/CT and PET/MR imaging. Furthermore, most data-driven motion estimation methods are better suited for motion detection and estimation of late PET frames in dynamic PET studies. The extension of these methods to earlier time points will be a focus in this study. The changing kinetics of the tracer distribution often complicate and prohibit motion estimation by various methods. Most motion estimation methods assume that the tracer distribution is (almost) stationary, such that apparent changes must be attributed to motion. Consequently, when the tracer distribution changes quickly, as it typically does early after injection, these estimation methods become less accurately or fail completely. An alternative strategy for this challenging problem could be to reconstruct also the attenuation image. Indeed, it has been shown that TOF-PET data are very rich, enabling not only the conventional reconstruction of the activity distribution, but also the joint estimation of the attenuation image. Because the attenuation must be stationary, the use of attenuation information may stabilize motion estimation. However, for an accurate reconstruction of the attenuation, the tracer distribution must satisfy some requirements, which are not necessarily met early after the tracer injection. Nevertheless, we believe that once the spread of activity is wider than the TOF resolution, anatomic (air cavities and bone) structures can provide useful information required for the spatial alignment.