Data-driven non-rigid motion detection and correction for NeuroEXPLORER
Jiawei Zhang, Chen Sun, Tommaso Volpi, Tieyong Zeng, Kathryn M. Fontaine, Yang Du, Takuya Toyonaga, John A. Onofrey, Yihuan Lu, Richard E. Carson · 2024
Patient movement in PET studies can lead to image blurring and inaccurate tracer concentration estimates. The next-generation brain PET system, NeuroEXPLORER (NX), has an integrated UIH markerless motion tracking system (UMT), which provides rigid head motion data by registering the upper facial surface. Additionally, the NX’s extended axial field of view increases sensitivity, aiding in extracting image-derived input functions (IDIF) from the carotid arteries (CA) and improving quantification in oncological studies. The lower face and neck areas, which include the CA or lesions, are prone to complex non-rigid motion patterns that cannot be expressed by rigid transformations and exceed UMT’s tracking capabilities. Therefore, effective MC is particularly vital for these areas. In this study, we introduced a data-driven non-rigid motion correction framework for the NX (NX-NRMC) to address non-rigid motions and assessed the effectiveness of UMT-derived rigid motion data in non-rigid regions. Our NX-NRMC, tested in two human studies, demonstrated promising results in improving quantification in non-rigid areas such as carotid arteries and tumors, areas where UMT-based rigid motion correction was inadequate. For the quantification of the tumor ROI, NX-NRMC yielded $\operatorname{SUV}_{\text {mean }}$ and $\operatorname{SUV}_{\text {max }}$ of 4.7 and 8.4, respectively, which were close to the motion-free reference values of 4.9 and 8.8. In the future, we will integrate the captured non-rigid motion estimates into an event-by-event motion correction reconstruction as well as combine both UMT-based rigid motion and NX-NRMC-based non-rigid motion to represent patient movement more accurately.