Nuklearmedizin 2019; 58(02): 133
DOI: 10.1055/s-0039-1683554
Vorträge
PET-MRT
Georg Thieme Verlag KG Stuttgart · New York

Continuous MR-based Motion Detection for Motion Correction in Hybrid PET/MRI

L Frohwein
1   Universität Münster, European Institute for Molecular Imaging, Münster
,
F Büther
2   Universitätsklinik Münster, Klinik für Nuklearmedizin, Münster
,
KP Schäfers
1   Universität Münster, European Institute for Molecular Imaging, Münster
› Institutsangaben
Weitere Informationen

Publikationsverlauf

Publikationsdatum:
27. März 2019 (online)

Auch verfügbar auf
 
 

    Ziel/Aim:

    We propose a real-time MRI (RT-MRI) sequence and processing scheme with which it is possible to track and correct motion during the complete PET scan duration and to make extensive use of motion information for PET motion correction.

    Methodik/Methods:

    RT-MRI data is acquired with a single-shot TurboFLASH sequence with a 35 ms slice acq. time (FoV of 400 × 400 × 180 mm3) with 20 slices using an interleaved slice scheme (first all even, then all odd slices). Respiratory signals for every slice position obtained using principal-component analysis (PCA) are used to first find slices with the correct resp. phase within the even subvolume and then supplement each subvolume with slices of the odd subvolume. This way, the sorting is constrained to real data by simultaneously doubling the volume sampling freq. Motion vector fields (MVF) between every time frame and the first ref. frame are determined using elastic registration. The scheme is validated using a dyn. thorax phantom.

    Ergebnisse/Results:

    For a 10-min. phantom scan, the volume resorting finds an avg. of 31 fitting slices/time point and slice pos. improving the temporal coherence (evaluated on a static phantom scan). Differences in liver dome lifts of up to 20 mm (max. amp. of phantom) are eliminated by the scheme. In areas with apparent motion (lung/heart), MVFs can be estimated smoothly in accordance to ground-truth motion of the phantom. Applied to the MRI data itself, motion is eliminated with no apparent artifacts. Correlation is increased from R = 0.83 for the non-corrected to R = 0.99 for the corrected end-insp. data (compared to static ref. in end-expir.).

    Schlussfolgerungen/Conclusions:

    RT MRI data is promising for an accurate MC of PET data as there is motion information available for every time point. Using real-time MRI for the motion correction may increase detection accuracy of small lesions in areas of motion, esp. for irregular motion such as bulk motion, baseline drift, etc.


    #