Optimization of a Regularization Weight for Bias Field Correction with Gray-level Co-occurrence Matrix in Image with PET/MR Imaging System

초록

The heterogeneity of images generated by positron emission tomography (PET)/magnetic resonance (MR)imaging systems significantly compromises diagnostic accuracy in the medical field. This study developed and refined a bias field correction strategy leveraging a gray-level co-occurrence matrix (GLCM) to enhance the uniformity of MR images within a PET/MR fusion imaging framework. We utilized a spherical phantom imbued with solutions of NaCl and NaCl+NiSO4 for image acquisition, employing T2-weighted turbo spin echo(TSE) and half-Fourier-acquired single-shot turbo spin echo techniques. The algorithm introduced for unifor-mity enhancement fine-tunes the contrast and energy metrics of the GLCM to identify an optimal lambdavalue. The application of this algorithm to MR images resulted in a marked improvement in percentage image uniformity (PIU), displaying superior characteristics relative to the uncorrected images. Specifically, the application of our algorithm to phantom images prepared with NaCl + NiSO4 and using the T2-weighted TSE technique yielded an average PIU of 97.72 %. In summary, we modeled a GLCM-based algorithm that can correct uniformity in MR images and confirmed the applicability of the proposed method in PET/MR systems.

키워드