How to identify and avoid MRI-PET Imaging artifacts: Challenges and potential solutions

Objectives A single institution experience on a wide range of clinical PET-MRI image artifacts are presented; ranging from inconsequential to severe cases, mimicking visceral organ metastasis. To identify and alleviate such artifacts, practical solutions and related Root-cause analysis are offered.

Methods PET-MRI has the potential to add exceptional anatomic resolution and soft-tissue contrast while lowering total patient radiation exposure. However, in addition to MRI motion artifacts due to cardiac, respiratory, and involuntary patient motion, which may lead to same technical registration limitations in PET-MRI as in PET-CT, numerous MR related new type imaging artifacts may occur; commons are illustrated with case examples. Simultaneous PET-MR imaging reduce registration errors providing accurate co-registered PET MR images, and shorten the ‘total acquisition times’ since many functional MR images are acquired during PET acquisition. For quantitative PET imaging, the γ-photon attenuation correction of the reconstructed data is a must. However, unlike in PET-CT, direct measurement of linear attenuation coefficients is not possible in integrated PET-MR systems. Since there is no simple relation between MR image intensity and attenuation coefficients, attenuation maps (µ-maps) can be estimated by segmenting MR images and assigning attenuation coefficients to the compartments. This limitation is usually addressed, by converting the MR information to linear attenuation coefficients. However, because the MR signal in the body relates to the tissues proton density instead of to γ-photon attenuation, this conversion is challenging. There are 3 types of MR based attenuation correction categories: a) segmentation based method, which segment the MR data into 3-4 tissue classes and assign uniform linear attenuation coefficients to each tissue class b) co-registered MR images and corresponding µ-maps are utilized for AC c) methods that create attenuation maps using PET emission data and MR anatomic information. The AC possibilities depend on the subject of study, and the requirements differ between clinical and preclinical imaging

Results Because of the short T2 relaxation times of bone tissue, standard MR sequences do not permit to delineate bone tissue only based on the intensity of single voxels. Although time consuming, ultrashort echo time (UTE) sequences allow us to create µ-maps utilizing bone signals. For clinical whole-body (WB) MRI studies bone detection utilizing current UTE sequences are quite time consuming, and since WB applications are less affected when bone tissue is not accurately discriminated from soft tissue in attenuation maps, for WB MR-based AC, segmentation based methods that do not account for the bone gained popularity, without compromising the quantification accuracy for regions that are not close to bone tissue. Dixon technique is more popular, and has also enabled differentiation between soft and adipose tissues. Soft-tissue and adipose-tissue decomposition is achieved using a 3-point Dixon-like decomposition. Predefined linear attenuation coefficients are assigned to classified voxels to generate MRI-based AC (µ) -maps.

Conclusions In addition spatial deformation inherent to certain MR sequences, as in Echo-planar imaging based sequences and inaccurate AC µ-map related PET-MRI image artifacts, the following list of conditions illustrated with case examples: - Artifacts related to non-PET optimized MRI coils, and their locations during imaging - Patient respiratory, cardiac motions and unintentional body movements. - Bowel motion related - MRI susceptibility image artifacts either from objects extrinsic or intrinsic to the patient, resulting in PET image artifacts. - MRI wrap and pulsation artifacts RESEARCH SUPPORT: None