Image-based Motion Correction for the Siemens hybrid-MR/BrainPET Scanner

Ziel/Aim:

Head motion can degrade accuracy of time activity curves (TAC) in quantitative neuroimaging; this becomes especially true for long PET acquisition protocols. Apart from misalignment of regions of interest, also a bias occurs due to mismatches between the subject attenuation map (AM) in a fixed reference position and the various head positions. Any mismatch causes a bias of the attenuation correction (AC), but also of the scatter correction (SC) which usually depends on the AM. Thus, motion correction (MC) is a prerequisite for quantification. Where no external tracking device is available, motion parameters can be estimated directly from PET images. We have implemented a robust MC workflow based on PET image rigid co-registration. Now, it can be routinely used for neuroimaging studies with our Siemens hybrid MR/BrainPET scanner [1].

Methodik/Methods:

Firstly, PET images are reconstructed according to the desired framing scheme, but without applying AC and SC. For each series of such (motion-uncorrected) images, we perform a co-registration with respect to a reference image using PMOD [2]. The found rigid transformations can be congruently applied to the AMs. Finally, images are reconstructed with PRESTO [3] and using the Multiple Acquisition Frame (MAF) method [4] with matched AMs.

Ergebnisse/Results:

We evaluated the method for [¹¹C]flumazenil as well as [¹¹C]ABP688 measurements with a series of subsequent frames of 2 – 5 minutes acquisition time (30 min. post-injection). The limited statistics due to the short frame length is still sufficient for the co-registration. After registration of the frames and applying the transformations according to MAF, any visible misalignment between images disappears and outliers in TACs are reduced.

Schlussfolgerungen/Conclusions:

The image-based head MC method allows to routinely detect and compensate inter-frame motion without additional tracking hardware. The accuracy of TACs can be evidently improved in case of subject motion. Intra-frame motion becomes of minor relevance due to the short frame length.

Literatur/References:

[1] IEEE TRPMS, Aug 2018, early access.

[2] http://www.pmod.com.

[3] IEEE TMI, 30, 2011, 879 – 892.

[4] IEEE TMI, 16, 1997, 137 – 144.