A modulated closed form solution for quantitative susceptibility mapping — A thorough evaluation and comparison to iterative methods based on edge prior knowledge

Abstract

The aim of this study is to perform a thorough comparison of quantitative susceptibility mapping (QSM) techniques and their dependence on the assumptions made. The compared methodologies were: two iterative single orientation methodologies minimizing the l2, l1TV norm of the prior knowledge of the edges of the object, one overdetermined multiple orientation method (COSMOS) and a newly proposed modulated closed-form solution (MCF). The performance of these methods was compared using a numerical phantom and in-vivo high resolution (0.65 mm isotropic) brain data acquired at 7 T using a new coil combination method. For all QSM methods, the relevant regularization and prior-knowledge parameters were systematically changed in order to evaluate the optimal reconstruction in the presence and absence of a ground truth. Additionally, the QSM contrast was compared to conventional gradient recalled echo (GRE) magnitude and R2* maps obtained from the same dataset. The QSM reconstruction results of the single orientation methods show comparable performance. The MCF method has the highest correlation (corr MCF = 0.95, r 2 MCF = 0.97) with the state of the art method (COSMOS) with additional advantage of extreme fast computation time. The L-curve method gave the visually most satisfactory balance between reduction of streaking artifacts and over-regularization with the latter being overemphasized when the using the COSMOS susceptibility maps as ground-truth. R2* and susceptibility maps, when calculated from the same datasets, although based on distinct features of the data, have a comparable ability to distinguish deep gray matter structures. © 2014 Elsevier Inc. All rights reserved. IntroductionPhase imaging has shown over the last decade to offer a good contrast, both between and within brain tissues in respect to the conventional magnitude signal (Duyn et al., 2007;Rauscher et al., 2005) as well as veins and iron rich regions (Haacke et al., 2004). The effect observed in the phase is known to be non-local, it reflects the magnetic field induced by the tissues' magnetic susceptibility (Marques and Bowtell, 2005), which scales linearly with the increase of the fields strength (making it suitable at high field strengths).Several studies have been performed on the origin of the susceptibility contrast with the main modulators being iron and myelin. Iron contributes to tissue contrast especially in the deep gray matter (globus pallidus, putamen and caudate) which has histologically derived high iron concentration showing good correlation with phase and susceptibility contrast (Bilgic et al., 2012;Schweser et al., 2011;Wharton and Bowtell, 2010). The other proposed contributor to the phase contrast, particularly between white and gray matter, is myelin where pathological demyelination has shown a decreased phase contrast between gray and white matter (C. Liu et al., 2011;Lodygensky et al., 2012) and good correlation was found between myelination and phase contrast during development. (Lodygensky et al., 2012).In addition to the no...

Studying cyto and myeloarchitecture of the human cortex at ultra-high field with quantitative imaging: R1, R2* and magnetic susceptibility

“…Other regions such as parietal cortex, and some parts of the motor, sensory cortex could suffer from such an effect in the gyrus, but not in the fissures and sulcus which will represent the majority of the cortical surface. In this manuscript, we took advantage of the high spatial resolution achievable at high field strengths (increasing the number of pixels between edge of brain mask and pial surface), and choose a background field removal technique, based in a preliminary study (Khabipova et al, 2015), that provided filtered fieldmaps whose susceptibility maps had the expected features shown in relaxation based cortical maps. It is nevertheless expectable that there would still be room for improvement as it is apparent in the QSM slice of Fig.…”

“…The subjects were removed from the scanner bore at the end of each scan, the subjects' head was re-positioned by the experimenter and was carefully cushioned to ensure comfort and minimise motion artefacts. A two step co-registration pipeline was used: first the magnitude images pre-brain extraction were co-registered (using a 12 parameters) to the normal head position; brain extraction was then applied and a final co-registration was then conducted (as detailed in Khabipova et al (2015)). The exact head rotations were determined by combining the co-registration matrices obtained using FLIRT (www.…”

“…The multi-echo GRE data from the different coils were combined using pixel by pixel SVD decomposition of the channel by echo time matrix, field maps were obtained by using pixel by pixel SNR weighting (Khabipova et al, 2015) and R 2 * map were calculated using signal, S, decay integration over the number of echo times, nTE,…”

How to choose the right MR sequence for your research question at 7 T and above?

“…The relaxation times found in the brain are summarised in table 1. 80-110 (i,j,k) 80-135 (i,j,k) a) (Bartha et al, 2002) b) (Koopmans et al, 2008) c) (Peters et al, 2007) d) (Pohmann et al, 2016) e) (Rooney et al, 2007) f) (Wright et al, 2008) g) (Yacoub et al, 2001) h) (Marques et al, 2010a) i) (Deistung et al, 2013) j) ( Khabipova et al, 2015) k) (Wharton and Bowtell, 2010) l) (Li et al, 2012) m) (Voelker et al, 2016) The experimenter is clearly limited to some degree not only by the characteristics of high field imaging, but also by the hardware available. At the time of writing the baseline performance for gradient systems is a strength of 40-80 mT/m and a slew rate of 200 mT/m s. The currently commercially available Siemens Magnetom Terra 7T can achieve 80mT/m and the GE M950 can achieve 50mT/m both at 200mT/m/ms.…”

Combining phase images from array coils using a short echo time reference scan (COMPOSER)

“…Building on phase difference imaging 18, a number of recent methods pose sophisticated, low‐noise solutions to the multi‐echo problem (eg, using singular valued decomposition 19 and the methods known as MAGPI 20 and CAMPUS 21). …”