Joint K-B Space Image Reconstruction and Data Fitting for Diffusion-Weighted Magnetic Resonance Imaging
Presentations
WE-A-201-4 (Wednesday, 7/13/2022) 7:30 AM - 8:30 AM [Eastern Time (GMT-4)]
Room 201
Purpose: Diffusion-weighted imaging (DWI) can provide clinically valuable quantitative information of Apparent Diffusion Coefficient (ADC). Conventional approach first reconstructs MR images at different b-values using corresponding k-space data, and then for each voxel, fits an exponential decay model to derive the ADC map. Considering the strong correlation of MR images along the spatial and b directions, this study proposes a joint k-b space reconstruction method to improve ADC accuracy by taking advantage of these correlations.
Methods: We formulated the reconstruction model as an optimization problem to simultaneously solve MR images at all b-values and the ADC map. These quantities were subject to a self-consistency condition of the exponential decay form. The objective function included a data fidelity term enforcing the agreement between the solution and the measured k-space data, and spatial regularization terms of the block-matching and 3D filtering(BM3D) denoising form on the MR images and ADC map. We solved the optimization problem using Alternating-Direction Method of Multipliers. We demonstrated effectiveness of our method in simulation cases and in phantom cases by adding Gaussian noise to the k-space data. We compared ADC map accuracy derived by our method with that obtained by the conventional approach.
Results: The joint reconstruction method was more tolerable to noise than the conventional method. In the simulation cases with added noise to yield b-value dependent SNR range of 9.9~21.8, the median relative error (MRE) of ADC map reconstructed by the conventional method was 6.1%, and our method reduced MRE to 1.1%. The improvement was mainly observed in area with intermediate SNR ratios. In the phantom cases with added noise to generate SNR range of 5.0~14.2, our method reduced MRE from 9.0% to 5.5%.
Conclusion: The proposed joint reconstruction method is effective in improving ADC map accuracy when fitting diffusion-weighted images with excessive noise level.
Keywords
Diffusion, Reconstruction, Noise Reduction
Taxonomy
Contact Email