Optimized 3D co-registration of ultra-low-field and high-field magnetic resonance images

Research output: Contribution to journalArticleScientificpeer-review

Researchers

  • Roberto Guidotti
  • Raffaele Sinibaldi
  • Cinzia De Luca
  • Allegra Conti
  • Risto Ilmoniemi

  • Koos Zevenhoven

  • Per E. Magnelind
  • Vittorio Pizzella
  • Cosimo Del Gratta
  • Gian Luca Romani
  • Stefania Della Penna

Research units

  • Gabriele d’Annunzio University
  • Los Alamos National Laboratory

Abstract

The prototypes of ultra-low-field (ULF) MRI scanners developed in recent years represent new, innovative, cost-effective and safer systems, which are suitable to be integrated in multi-modal (Magnetoencephalography and MRI) devices. Integrated ULF-MRI and MEG scanners could represent an ideal solution to obtain functional (MEG) and anatomical (ULF MRI) information in the same environment, without errors that may limit source reconstruction accuracy. However, the low resolution and signal-to-noise ratio (SNR) of ULF images, as well as their limited coverage, do not generally allow for the construction of an accurate individual volume conductor model suitable for MEG localization. Thus, for practical usage, a high-field (HF) MRI image is also acquired, and the HF-MRI images are co-registered to the ULF-MRI ones. We address here this issue through an optimized pipeline (SWIM—Sliding WIndow grouping supporting Mutual information). The co-registration is performed by an affine transformation, the parameters of which are estimated using Normalized Mutual Information as the cost function, and Adaptive Simulated Annealing as the minimization algorithm. The sub-voxel resolution of the ULF images is handled by a sliding-window approach applying multiple grouping strategies to down-sample HF MRI to the ULF-MRI resolution. The pipeline has been tested on phantom and real data from different ULF-MRI devices, and comparison with well-known toolboxes for fMRI analysis has been performed. Our pipeline always outperformed the fMRI toolboxes (FSL and SPM). The HF–ULF MRI co-registration obtained by means of our pipeline could lead to an effective integration of ULF MRI with MEG, with the aim of improving localization accuracy, but also to help exploit ULF MRI in tumor imaging.

Details

Original languageEnglish
Article numbere0193890
Pages (from-to)1-19
JournalPloS one
Volume13
Issue number3
Publication statusPublished - 1 Mar 2018
MoE publication typeA1 Journal article-refereed

Download statistics

No data available

ID: 18276173