Quality control protocol for frame-to-frame PET motion correction

Subject motion during Position Emission Tomography (PET) brain scans can reduce image quality, and may lead to incorrect biological outcome measures, especially during analysis of dynamic data sets. This is particularly relevant when imaging with state-of-the-art scanners such as the High Resolution Research Tomograph (HRRT, Siemens Medical Solutions). Motion correction via frame-to-frame image realignment is simpler to implement and requires fewer computing resources than methods that correct for motion during data reconstruction and has been shown to significantly improve the accuracy of dynamically-derived biological variables. However, an ongoing problem is a lack of objective criteria to validate the accuracy of frame-to-frame realignment. Visual inspection of realigned images is a common method of validation but requires a significant amount of operator time and results may vary from one operator to another. This work presents a quality control protocol that automatically flags inadequate realignments based on the comparison of motion transformation matrices obtained from two independent sources: the Polaris Vicra optical tracking device and the image based realignment algorithm AIR (Automated Image Registration). A metric was computed to determine the difference between the transformations from both methods. Realignments were accepted or flagged based on the value of the metric. Since the two methods rely on independent motion assessment tools, the chance of both algorithms giving consistently wrong estimates is low. Human test cases show that the quality control protocol is capable of correctly identifying both acceptable and incorrect realigned images, thus providing an objective quality control metric. Implementation of the protocol reduces the number of images requiring visual inspection by 72% and operator time required by 50%, decreasing both operator labour and operator-dependent biases.