Image derived input function in PET brain studies: Blood-based methods are insensitive to motion artifacts

Objectives A recent study using a blood-free image-derived input function (IDIF) method from carotid arteries showed that IDIF is very vulnerable to patient’s motion. The aim of this study was to test, in both phantom and clinical [¹¹C](R)-rolipram studies, the robustness of a blood-based IDIF method with regards to motion artifacts.

Methods First, the impact of motion on IDIF accuracy was assessed with an analytical simulation with realistic noise and resolution properties of the HRRT using a numerical phantom of the human brain, equipped with internal carotids. Different degrees of translational (from 1 to 20 mm) and rotational (from 1 to 15°) motions were tested. Then the impact of motion was tested with dynamic scans of 3 healthy volunteers acquired on the HRRT, reconstructed with and without an on-line motion correction system. IDIFs and Logan-VT values derived from simulated and clinical scans without motion correction were compared to those obtained in the scans with motion correction.

Results In phantoms, difference of area under the carotid time activity curve (AUC) with motion as compared to that without motion was up to 19% for rotations and 66% for translations. However, for the final partial-volume corrected IDIFs, which are fitted to blood samples, the AUC difference was only up to 11% for rotations and 8% for translations. Logan-VT errors were always <10%, except for the maximal translation of 20 mm, in which the error was 18%. The amount of movement in the clinical scans was less important than that simulated, with a maximum average translation and motion of about 11 mm and 11°, respectively. Errors in the clinical scans without motion correction appears to be insignificant, with differences in AUC and Logan-VT always <10% as compared to scans with motion correction.

Conclusions When a blood-based IDIF method is used, patient’s motion has only a minor impact on IDIF estimation and brain kinetic modeling