Motivation: FSE MRI enables rapid scan times, but the reduced amount of k-space data acquired makes it highly susceptible to motion artifacts, complicating effective motion correction. Goal(s): Develop a comprehensive motion correction framework for FSE MRI that operates in k-space, preserving structural fidelity and reducing scan time. Approach: A hybrid model combining Diffusion and DIP networks was used to detect and correct motion-affected TRs in k-space, integrating L2, HaarPSI, and VIF losses for improved quality. Results: The method effectively corrected motion artifacts, yielding higher-quality MRI images by blending clean and corrected k-space data. Impact: This framework provides robust motion correction for MRI, preserving details and enabling broader applications for complex-domain MRI data.
Son et al. (Tue,) studied this question.