Premium
Methods for the computation of templates from quantitative magnetic susceptibility maps (QSM): Toward improved atlas‐ and voxel‐based analyses (VBA)
Author(s) -
Hanspach Jannis,
Dwyer Michael G.,
Bergsland Niels P.,
Feng Xiang,
Hagemeier Jesper,
Bertolino Nicola,
Polak Paul,
Reichenbach Jürgen R.,
Zivadinov Robert,
Schweser Ferdinand
Publication year - 2017
Publication title -
journal of magnetic resonance imaging
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.563
H-Index - 160
eISSN - 1522-2586
pISSN - 1053-1807
DOI - 10.1002/jmri.25671
Subject(s) - template , voxel , computer science , quantitative susceptibility mapping , artificial intelligence , modality (human–computer interaction) , contrast (vision) , atlas (anatomy) , magnetic susceptibility , pattern recognition (psychology) , magnetic resonance imaging , physics , medicine , quantum mechanics , radiology , programming language , anatomy
Purpose To develop and assess a method for the creation of templates for voxel‐based analysis (VBA) and atlas‐based approaches using quantitative magnetic susceptibility mapping (QSM). Materials and Methods We studied four strategies for the creation of magnetic susceptibility brain templates, derived as successive extensions of the conventional template generation (CONV) based on only T 1 ‐weighted ( T 1 w) images. One method that used only T 1 w images involved a minor improvement of CONV (U‐CONV). One method used only magnetic susceptibility maps as input for template generation (DIRECT), and the other two used a linear combination of susceptibility and T 1 w images (HYBRID) and an algorithm that directly used both image modalities (MULTI), respectively. The strategies were evaluated in a group of N = 10 healthy human subjects and semiquantitatively assessed by three experienced raters. Template quality was compared statistically via worth estimates (WEs) obtained with a log‐linear Bradley‐Terry model. Results The overall quality of the templates was better for strategies including both susceptibility and T 1 w contrast (MULTI: WE = 0.62; HYBRID: WE = 0.21), but the best method depended on the anatomical region of interest. While methods using only one modality resulted in lower WEs, lowest overall WEs were obtained when only T 1 w images were used (DIRECT: WE = 0.12; U‐CONV: WE = 0.05). Conclusion Template generation strategies that employ only magnetic susceptibility contrast or both magnetic susceptibility and T 1 w contrast produce templates with the highest quality. The optimal approach depends on the anatomical structures of interest. The established approach of using only T 1 w images (CONV) results in reduced image quality compared to all other approaches studied. Level of Evidence: 2 Technical Efficacy: Stage 1 J. Magn. Reson. Imaging 2017;46:1474–1484.