Neuromelanin-MRI using 2D GRE and deep learning: considerations for improving the visualization of substantia nigra and locus coeruleus

TL;DR

This paper presents an optimized neuromelanin-MRI protocol using 2D GRE and deep learning to enhance visualization of the substantia nigra and locus coeruleus in a clinically feasible time frame.

Contribution

It introduces a combined optimization of imaging parameters and deep learning reconstruction for improved SN and LC visualization in neuromelanin-MRI.

Findings

High-resolution imaging with deep learning denoising improves SN and LC depiction.

The optimized protocol is feasible within a 7-minute scan time.

Deep learning enhances image quality and visualization accuracy.

Abstract

An optimized clinically feasible neuromelanin-MRI imaging protocol for visualising the SN and LC simultaneously using deep learning reconstruction is presented. We optimize flip-angle for optimal combined SN and LC depiction. We also experimented with combinations of anisotropic and isotropic in-plane resolution, partial vs full echoes and the number of averages. Phantom and in-vivo experiments on three healthy volunteers illustrate that high-resolution imaging combined with deep-learning denoising shows good depiction of the SN and LC with a clinically feasible sequence of around 7 minutes.