Image space identification of a motion tracking tool in PET and PET/CT

Subject motion during PET data acquisition reduces the achievable resolution. Methods to correct PET data for motion often use external motion tracking devices, for example with commercial Infrared based cameras. Accurate spatial and temporal alignment between the Infrared camera and PET or PET/CT is essential. Here we propose a simple and robust method for direct identification in PET image space of the motion tracking tool that is attached to the head to calculate the ground truth transformation between PET and IR tracking space. This is achieved by identifying the set of reflective markers of the tracking tool in the attenuation correction image from either the conventional transmission scan of the PET only system or the CT scan of the combined PET/CT system. The tool detection process can be complicated however, with careful design of the tracking tool, it can be simple. The tool detection consists of applying a threshold to the attenuation images to select only the highest attenuating regions and using blob detection techniques to get coordinates. These regions can be designed to correspond with the positions of the reflective markers. Importantly, due to an extension of the small angles approximation, the accuracy of the tool detection seems to have little effect on the overall motion correction transformation. By mis-calculating the tool position by 5 mm the resultant effect changes a voxel position by 0.092 mm.