Temporal Regularization Use in Dynamic Contrast-Enhanced MRI.

[摘要] Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) studies demand both high spatial and high temporal resolution. We need high spatial resolution to accurately visualize tissue morphology, and we need high temporal resolution to accurately follow the contrast kinetics of the tissue, which provide clinically important physiological information. We can only acquire so many measurements per unit time, however, and this limited data implies an inherent tradeoff between spatial and temporal resolution in the reconstructed image sequence. Most existing methods undersample the data and then employ some sort of data sharing technique in the k-space domain to recover the ;;missing’ data points. These data sharing schemes are based on an implicit assumption that the dynamic object varies smoothly in time. We present an image reconstruction scheme based on an object domain model that does not attempt any k-space data recovery, but rather explicitly uses the assumption of temporal smoothness in the image domain to estimate the image sequence that best fits the available data. Our proposed method is called Temporal Regularization Use in Image Reconstruction (TRUIR), and is a penalized likelihood formulation that includes spatial and temporal regularization terms in addition to the data fidelity term. This work presents our TRUIR formulation for both single coil and parallel imaging, and explores various aspects of TRUIR reconstructed image sequences. We evaluate the effect of the spatial and temporal regularization parameters on the resolution properties of TRUIR reconstructed image sequences, and present our work toward establishing selection criteria for these parameters. In evaluating our proposed TRUIR method, we focus on the application of DCE-MRI in the characterization and assessment of breast cancer. Our simulation studies model contrast uptake in a dynamic digital breast phantom. Results show that TRUIR reconstructions offer improved temporal dynamics when compared to more traditional frame-by-frame reconstructions, as well as more accurate estimates of kinetic parameters, particularly when TRUIR is used in conjunction with two new proposed k-space sampling trajectories, which are also presented in this work. These new trajectories are also shown to be more robust to regularization parameter choice. Further work is needed to improve TRUIR’s spatial resolution.

[发布机构] University of Michigan

[效力级别] Regularization

[关键词] Image Reconstruction;Regularization;DCE-MRI;Breast Cancer;Biomedical Engineering;Electrical Engineering;Engineering;Biomedical Engineering