A fiducial marker based technique for alignment of simultaneously acquired PET and MRI images

Combined MRI and PET systems are currently being developed by a number of research groups and also commercially. Most systems have been designed to permit the simultaneous acquisition of MRI and PET data. The approach taken by many groups, is to build an MRI compatible PET insert, that works inside a standard MRI scanner. Unlike other multi-modality imaging systems such as PET/CT, the physical location of the PET scanner within the MRI scanner may vary each time the PET scanner is removed and replaced in the MRI scanner. In order to produce fully aligned PET and MRI images over the same region of the subject, the PET scanner location within the MRI field of view (FOV) is required, as well as the transformation between the PET and the MRI images. We have developed such a technique for our single slice pre-clinical MR-compatible PET system that uses long optical fibres to distance the PET PMTs from the high field at the centre of the MRI scanner. The method uses MRI visible markers attached to the PET scanner at a known position with respect to the PET FOV. In our configuration the markers must be positioned close to MRI compatible PET gamma shields which are used to improve the signal to noise ratio of the PET images by reducing the number of events recorded from activity outside the FOV. These shields are made from BGO and PbO, which have similar magnetic properties to LSO. The main magnetic field of the MRI scanner becomes distorted near to the shields, although they impose little distortion over the useful FOV of the PET system. We have assessed the accuracy with which the MRI visible markers can be used to select the location of the PET imaging slice and to align the PET and MRI images in-plane, using simultaneously acquired data of various phantoms. In order to carry out the in-plane registration, two separate MRI images had to be acquired in which the direction of the MRI frequency-encoding gradient was flipped between the two acquisitions to estimate the in-pl- - ane shift in marker location caused by field inhomogeneities imposed by the PET shielding materials. The slice location accuracy and the in-plane registration appear to be accurate to approximately 0.5 mm. We have successfully applied the method to produce simultaneously acquired, registered ¹⁸F^-PET and MRI images of the mouse neck. A similar method is likely to be required for many future pre-clinical simultaneous PET and MRI studies as it will be essential for situations in which there is little similarity between the PET and MRI images.