Abstract In magnetic resonance imaging (MRI), variations in scan parameters and scanner specifications can result in differences in image appearance. To minimize these differences, harmonization in MRI has been suggested as a crucial image processing technique. In this study, we developed an MR physics-based harmonization framework, Physics-Constrained Deep Neural Network for Multisite and multiscanner Harmonization (PhyCHarm). PhyCHarm includes two deep neural networks: (1) the Quantitative Maps Generator to generate T 1 - and M 0 -maps and (2) the Harmonization Network. We used an open dataset consisting of 3T MP2RAGE images from 50 healthy individuals for the Quantitative Maps Generator and a traveling dataset consisting of 3T T 1 w images from 9 healthy individuals for the Harmonization Network. PhyCHarm was evaluated using the structural similarity index measure (SSIM), peak signal-to-noise ratio (PSNR), and normalized-root-mean square error (NRMSE) for the Quantitative Maps Generator, and using SSIM, PSNR, and volumetric analysis for the Harmonization network, respectively. PhyCHarm demonstrated increased SSIM and PSNR, the highest Dice score in the FSL FAST segmentation results for gray and white matter compared to U-Net, Pix2Pix, and CALAMITI. PhyCHarm showed a greater reduction in volume differences after harmonization for gray and white matter than U-Net, Pix2Pix, or CALAMITI. As an initial step toward developing advanced harmonization techniques, we investigated the applicability of physics-based constraints within a supervised training strategy. The proposed physics constraints could be integrated with unsupervised methods, paving the way for more sophisticated harmonization qualities.

Quantitative magnetization transfer (qMT) imaging is sensitive to myelin-related macromolecular content and brain microstructure but is limited by long scan times. We present a fast, SAR-efficient qMT technique using a segmented echo-planar imaging readout with variable power MT preparation (EP-vpMT). EP-vpMT was implemented at 3T (1.5×1.5×4.0 mm³ voxels; 9 µL) using a 3D segmented EPI readout with modulated MT RF pulses to reduce SAR while preserving contrast. Pseudo bound pool fraction (pseudo-BPF) maps were obtained from healthy participants. Consistency with pseudo-BPF derived from conventional GRE-MT and repeatability was subject to Bland-Altman analysis. Multiple sclerosis (MS) patients were examined at 3T and at 7T (2.0 mm isotropic voxels; 8 µL) to explore feasibility for assessing tissue integrity and for application at ultra-high field. EP-vpMT achieved whole-brain qMT in 6 min 25 sec, reducing scan time by 76% compared to GRE-based qMT (26 min 20 sec) while maintaining similar SAR levels. Strong agreement was observed between methods, and test-retest reliability showed minimal bias with 95% limits of agreement within a clinically negligible range. In MS patients, EP-vpMT delineated lesions at 3T and at 7T. EP-vpMT enables fast qMT imaging at 3T with strong agreement with conventional methods. Its ability to detect MS lesions and to translate to ultra-high field MRI supports future use for assessing myelin-related macromolecular content.