Frame misalignment-induced errors in PET studies: an investigation on strategies for correction

The viability of a PET study relies on the patient´s ability to stay still during the lengthy data acquisition process. Patient motions during exam can corrupt the data analysis in two ways: 1) they may alter ROI-based analysis of dynamic data due to inter-frame shifts. 2) they systematically introduce errors due to emission - transmission data mismatches. In typical data analysis processes, frame alignment is usually performed via registration algorithms which aim at computing the optimal spatial transformation between a target and each frame of the PET volume. Within the scope of developing a strategy for systematically correcting dynamic brain volumes, we carried out investigations using simulated data, whose interests were twofold 1) to evaluate separately the consequences of such effects in typical brain studies and 2) to assess different correction algorithms. From these studies, we conclude than : emission transmission mismatches are major sources of errors by distorting the measured tracer distributions entailing miss-correction in the frame alignment process. Finally, these investigations showed the importance of using external data such as MRI-based volumes for target definition.