Joint Optimization of Sampling Pattern and Image Reconstruction for 3D Dual-Echo Spoiled Gradient Echo MRI

Motivation: Conventional 3D dual-echo spoiled gradient echo MRI provides valuable T1 contrast and water-fat separation, but long scan times hinder clinical applications. Goal(s): This work aims to accelerate image acquisition and improve image quality for dual-echo MRI by optimization of both k-space sampling and image reconstruction. Approach: We extend the AutoSamp framework to jointly optimize sampling patterns and reconstruction for dual-echo MRI using a data-driven approach based on variational information maximization. Results: Our method achieved significantly enhanced image quality in accelerated dual-echo imaging, demonstrating improved reconstruction accuracy. Impact: This work offers a novel approach to accelerating dual-echo MRI by enhancing image quality through joint sampling and reconstruction optimization. Our method provides valuable insights into sampling pattern design and reconstruction strategies, potentially broadening clinical applications of fast multi-echo MRI.