End-to-end Complex Image Reconstruction Pipeline for GE Diffusion MRI Acquisitions

Motivation: Reconstructing diffusion MRI (dMRI) data in the complex domain, which provides superior denoising capabilities compared to magnitude data, is not available as standard from some scanner manufacturers. Goal(s): Our goal was to create a scalable offline reconstruction pipeline for dMRI acquisitions on GE HealthCare scanners to access complex diffusion data. Approach: We developed an end-to-end Python pipeline for complex reconstruction by building upon the C++ Orchestra SDK available through GE, including capabilities for in-plane/out-of-plane accelerations. Results: We validated the pipeline by comparing scanner and offline magnitude reconstruction. We make the complex offline dMRI reconstruction available for GE scanners users. Impact: The offline reconstruction pipeline allows access to complex dMRI data, retrospectively switching filters off and supporting in-plane/out-of-plane accelerations and partial Fourier. This allows seamless operations in the complex domain (e.g. denoising) and potentially paves the way for vendor-agnostic image reconstruction.