Disentangling contributions from iron and myelin architecture to brain tissue magnetic susceptibility by using Quantitative Susceptibility Mapping ( QSM )


Journal article


F. Schweser, A. Deistung, K. Sommer, J. Reichenbach
2011

Semantic Scholar
Cite

Cite

APA
Schweser, F., Deistung, A., Sommer, K., & Reichenbach, J. (2011). Disentangling contributions from iron and myelin architecture to brain tissue magnetic susceptibility by using Quantitative Susceptibility Mapping ( QSM ).

Chicago/Turabian
Schweser, F., A. Deistung, K. Sommer, and J. Reichenbach. “Disentangling Contributions from Iron and Myelin Architecture to Brain Tissue Magnetic Susceptibility by Using Quantitative Susceptibility Mapping ( QSM )” (2011).

MLA
Schweser, F., et al. Disentangling Contributions from Iron and Myelin Architecture to Brain Tissue Magnetic Susceptibility by Using Quantitative Susceptibility Mapping ( QSM ). 2011.


Abstract

INTRODUCTION – Magnetic susceptibility is an intrinsic physical tissue property which recently became accessible in vivo by a novel imaging technique called quantitative susceptibility mapping (QSM) [1,2]. Susceptibility maps of the human brain demonstrate astounding anatomical contrast [1,2], which is currently believed to be predominantly due to iron (paramagnetic) and myelin-lipids (diamagnetic) [3]. The intermixing of both contributions, however, complicates interpretation of susceptibility changes in particular in neurodegenerative diseases where inflammatory myelin-loss and focal iron accumulation may occur simultaneously. It has, furthermore, recently been discovered that a considerable orientation dependence of brain tissue susceptibility exists, which further complicates interpretation. We present a novel technique for substantially increasing the specificity of QSM by utilizing additional R2* information. The technique yields two novel contrasts, one that is independent of orientation effects, whereas the other is independent of tissue iron concentration. THEORY – A three compartment tissue model was assumed with punctuate particle inclusions (called iron in the following) and myelinated axons in a homogenous tissue matrix. In this model, the bulk voxel susceptibility can be expressed by Eq. 1 (volume fraction of iron neglected) [1]. The corresponding relation for the effective transverse relaxation rate, R2*, is given by Eq. 2 [4,5]. In both equations the terms associated with myelin depend on the orientation of the axons relative to the main magnetic field (angle θ ) as described by Eq. 3 [5] and Eq. 4 [6,7]. The dependence on iron concentration may be eliminated from the equations by linear combination of Eq. 1 and Eq. 2 according to Eq. 5, yielding a novel iron-independent contrast ξnoFe. The coefficient 1 ˆ ˆ − Fe Fer χ may be estimated from literature values (this study; see Tab. 2) or from R2* and susceptibility values in regions with a similar contribution of myelin. The contrast ξnoFe depends linearly on the myelin-lipid volume fraction and includes the myelin-related orientation dependencies. A rotation invariant contrast may be generated by a linear combination of Eqs. 1 and 2 that eliminates the θ -terms according to Eq. 6. This contrast is linear with respect to both the iron concentration and the myelin volume fraction. MATERIALS AND METHODS – To demonstrate the technique, high-resolution double-echo GRE data was acquired from the brain of a volunteer (male, 26y) using the ToF-SWI-sequence [9] (TE1/TE2=3.38ms/22ms, TR=30ms, FA=20°, 600μm isotropic voxels; acquisition time: 15min.) on a 3 Tesla whole-body MRI scanner (Tim Trio, Siemens Medical Solutions, Erlangen, Germany) using a 12-channel receive head-matrix coil. The scan was repeated with the volunteer’s head in head-to-neck position to investigate orientation effects. The resulting complex-valued images were registered to the normal head position using FSL-FLIRT (FMRIB, Oxford University). R2* maps were computed from the magnitude echoes with compensation of Rician noise [10] and susceptibility maps were reconstructed from the phase images using the HEIDI algorithm [submitted to ISMRM]. The maps were, finally, combined according to Eqs. 5 and 6. The unknown constant 1 || ˆ − ⊥ ⋅ My r χ in Eq. 6 was determined by minimizing the difference between the orientation independent contrasts, ξnoOrient, of the two head orientations (A,B) in the corpus callosum: 2