A fully open-source framework for streaming and cloud-processing of low-field MRI data

TL;DR

This paper introduces an open-source framework enabling real-time streaming and cloud-based processing of low-field MRI data, facilitating advanced reconstruction and post-processing without high-performance hardware.

Contribution

It presents a novel open-source system that integrates scanner control with cloud processing, allowing scalable and portable low-field MRI workflows.

Findings

Reliable streaming performance demonstrated in experiments

Supports deep-learning-based denoising and distortion correction

Enables continuous acquisition-to-reconstruction workflows

Abstract

Purpose: To present a fully open-source framework for quasi-real-time streaming and cloud-based processing of low-field (LF) MRI data, addressing the growing computational demands of advanced reconstruction and post-processing pipelines in portable and affordable MRI systems. Methods: The proposed framework integrates open-source scanner control software with a network-enabled streaming architecture, allowing for raw data to be transmitted directly from the MRI console to remote compute resources. Cloud-based processing modules support image reconstruction and advanced post-processing, including computationally intensive physics- and learning-based methods, while maintaining compatibility with low-cost on-device control hardware. Results: The system enables continuous acquisition-to-reconstruction workflows in LF-MRI without requiring specialized high-performance console architectures. Selected example applications include deep-learning-based denoising, field-induced distortion correction, and non-Cartesian image reconstruction. Experimental demonstrations show reliable streaming performance. Conclusions: Open-source streaming and cloud-processing provide an effective pathway to overcome the computational limitations of embedded LF-MRI consoles. By decoupling acquisition hardware from intensive reconstruction workloads, the proposed framework supports scalable deployment of advanced algorithms while preserving the affordability and portability that motivate LF-MRI.