A Practical Quality Assurance Procedure for Data Acquisitions in Preclinical Simultaneous PET/MR Systems

Abstract

The availability of preclinical simultaneous PET/MR imaging systems has been increasing in recent years. Therefore, this technique is progressively moving from the hands of pure physicists towards those of scientists more involved in pharmacology and biology. Unfortunately, these combined scanners can be prone to artefacts and deviation of their characteristics under the influence of external factors or mutual interference between subsystems. This may compromise the image quality as well as the quantitative aspects of PET and MR data. Hence, quality assurance is crucial to avoid loss of animals and experiments. A possible risk to the acceptance of quality control by preclinical teams is that the complexity and duration of this quality control are increased by the addition of MR and PET tests. To avoid this issue, we have selected over the past 5 years, simple tests that can be easily and quickly performed each day before starting an animal PET/MR acquisition. These tests can be performed by the person in charge of the experiment even if this person has a limited expertise in instrumentation and performance evaluation. In addition to these daily tests, other tests are suggested for an advanced system follow-up at a lower frequency. In the present paper, the proposed tests are sorted by periodicity from daily to annual. Besides, we have selected test materials that are available at moderate cost either commercially or through 3D printing.

Keywords: MRI; PET; Preclinical imaging; Quality control; Simultaneous imaging.

Fig. 1

MR magnitude noise spectrum (kHz) obtained with a rat body quadrature coil in reception mode. (a) Unexpected peaks are visible, mainly at -18.5 kHz. This was caused by an unshielded high-definition multimedia interface wire in the imaging room. This unwanted peak will be materialised on the image in the form of white streaks artefacts particularly visible with gradient echo sequences. (b) Same acquisition after the cable was unplugged

Fig. 2

Picture of the mouse-sized cylinder phantom (on the right side) used for several PET and MR tests. This phantom is here presented near a rat-sized version (on the left side). The phantoms are 10-cm long to fit our PET axial FOV. The two phantoms have a diameter of 2 cm and 4 cm respectively for the mouse and rat

Fig. 3

Example of multiplanar GRE scout views of the mouse-sized cylindrical phantom acquired in 15 s. (a), (b), (c) The transaxial, coronal, and sagittal planes. The visible deformation is caused by a non-MR compatible 27-G catheter voluntary introduced in the FOV where the caudal vein would have been located during an in vivo acquisition

Fig. 4

Illustration of the SNR measurements. The green circular ROI is dedicated to the signal measurement, whereas the white rectangular ROIs, drawn outside the ghosting region, are used to measure the noise. Any artefact is reported

Fig. 5

GRE transverse scouts acquired with the rat-sized homogeneous cylinder centred in the rat RF coil. (a) Normal image aspect. (b, c) Line artefacts (marked with a red arrow) caused by electromagnetic interferences. (d) Signal heterogeneities caused by damages on the RF coil to preamplifier cables. Artefact and signal issues were solved by the optimization of electromagnetic shielding and improvement in the experimental procedure

Fig. 6

(a) Photography of the ultra-micro hot spot phantom used for the spatial resolution assessment displayed with the rods diameter in mm. (b) PET transaxial image of the ultra-micro hot spot phantom. The acquisition is 30 min long, and data are reconstructed with the OSEM algorithm using ten iterations and 64 subsets. In these conditions, the 0.8-mm rods are discernible

Fig. 7

MR images (displayed in colour) of the grid phantom here registered with an X-ray micro-CT scan (displayed in grey levels). (a) Transaxial view. (b) Coronal view, where a slight distortion is visible near the MR FOV boundaries. (c) Schematic representation of the registration of the CT and MR images of the grid phantom in which the arrows indicate the location of the typical misalignments observed between the MR images and the CT taken as reference

Fig. 8

(a) An example of transverse B0 distortion maps (ppm) obtained in our 7-Tesla magnet. (b) The effect of a non-MR compatible catheter voluntarily introduced in the FOV. The acquisition procedure remains unchanged. (c) A linear horizontal profile traced on the (a) and (b) maps. Curve number 2 shows the B0 distortion induced by the presence of the catheter compared to curve number 1 obtained after removing it

Fig. 9

(a) A picture of the NEMA IQ phantom. (b) The recovery coefficients insert of the phantom. (c) The homogeneity insert of the phantom. (d) The spill-over ratio region

Fig. 10

(a) An example of energy spectrum (counts per keV) acquired with our system. The measured energy resolution is about 20% with a 1-cm deep LYSO-CE crystal. (b) A dual-layer crystal response map normalised to the detector maximum count, showing acceptable heterogeneous areas. (c), (d) Non-compliant results obtained on the same detector due to an electric supply failure of an SiPM tile. A system reset fixed the issue

Fig. 11

Overview of the PET/MR image registration accuracy test object used in our facility. The object is made of three 1-mL syringes filled with 10 MBq of ¹⁸F. Simultaneously, acquired PET and MR datasets are exported in DICOM before being displayed in VivoQuant 2.0 (Invicro, Needham, MA, USA). MR images were resampled to the PET isotropic voxel size of 0.56 mm. (a) The maximum intensity projection (MIP) of the MR image (axial FOV of 8 cm). (b) The MIP of the PET image (axial FOV of 10 cm). (c) The MIP of the PET/MR registration