Evaluation of the Non-quadratic Model-based Reconstruction Method for Geometrical Distortion Correction in MR Imaging using Pari Head QC Phantom

A major issue when using MR images in precision radiotherapy planning is the geometrical distortion. The magnetic susceptibilities are the primary cause of the geometrical distortion in MR images which could be up to 4 mm across air cavities and body tissue boundaries. The conventional distortion correction approach is not able to completely eliminate the effect of the geometrical distortions. To address this issue, a model-based reconstruction is proposed which benefits from two regularization terms. A regularization term is employed for the correction in the field-map and the other one in the image domain. Regarding the edge smoothing effect of the quadratic functions, non-quadratic penalty functions are proposed for the model-based image reconstruction. In this framework, optimization of the parameters for non-quadratic regularization is highly challenging. In this light, we proposed an approximation approach to select optimized parameters for the model-based reconstitution method. An advantage of the proposed method is that it does not require prior knowledge regarding the ground truth data. To validate the proposed method, BrainWeb phantom was employed for the simulation study. The effectiveness of the proposed approach was further demonstrated experimentally using the Pari head QC phantom customized by SpinTecx-LLC. This phantom would simulate magnetic susceptibility effects in the image due to the presence of air cavities. The simulation study showed a good agreement between the results of the proposed method and the ground truth data. The phantom study further demonstrated the superiority of the proposed method over the conventional approach, wherein the maximum relative intensity difference with respect to the uncorrected image was 10.0% and 2.5% for the proposed approach and the conventional method, respectively. The signal recovery was remarkable for the edges in the non-homogenous regions.