Gray matter myelination of 1555 human brains using partial volume corrected MRI images

Highlights

• Without partial volume correction T1w/T2w myelination may be overestimated.

• Average myelin maps presented for 1555 clinically normal 18–35 year old subjects.

• Inner cortical layers show higher T1w/T2w myelination compared to outer layers.

• T1w/T2w myelination of the cortex increases with age between 18 and 35 years.

• Age-dependent increase of myelination occurs mostly in the inner cortical layers.

Abstract

The myelin content of the cortex changes over the human lifetime and aberrant cortical myelination is associated with diseases such as schizophrenia and multiple sclerosis. Recently magnetic resonance imaging (MRI) techniques have shown potential in differentiating between myeloarchitectonically distinct cortical regions in vivo. Here we introduce a new algorithm for correcting partial volume effects present in mm-scale MRI images which was used to investigate the myelination pattern of the cerebral cortex in 1555 clinically normal subjects using the ratio of T1-weighted (T1w) and T2-weighted (T2w) MRI images. A significant linear cross-sectional age increase in T1w/T2w estimated myelin was detected across an 18 to 35 year age span (highest value of ~ 1%/year compared to mean T1w/T2w myelin value at 18 years). The cortex was divided at mid-thickness and the value of T1w/T2w myelin calculated for the inner and outer layers separately. The increase in T1w/T2w estimated myelin occurs predominantly in the inner layer for most cortical regions. The ratio of the inner and outer layer T1w/T2w myelin was further validated using high-resolution in vivo MRI scans and also a high-resolution MRI scan of a postmortem brain. Additionally, the relationships between cortical thickness, curvature and T1w/T2w estimated myelin were found to be significant, although the relationships varied across the cortex. We discuss these observations as well as limitations of using the T1w/T2w ratio as an estimate of cortical myelin.

Introduction

Myelin expedites the conduction of electrical signals along axons and is essential for normal, healthy function of the nervous system. While most abundant in the cerebral and cerebellar white matter, significant amounts of myelinated fibers are present in the cortical gray matter. Classic histological studies of postmortem brains by Vogt, Campbell, and Elliot Smith (Nieuwenhuys, 2013) depict how the distribution of myelinated fibers can vary significantly between cortical regions. Recently, advances of MRI techniques enabled the investigation of the estimated myelin content of the human brain in vivo. Studies have shown that the T1, T2 and T2* relaxation times of the brain tissue depend on the myelin content of the tissue (Bock et al., 2009, Bock et al., 2011, Bock et al., 2013, Geyer et al., 2011, Clark et al., 1992, Barbier et al., 2002, Walters et al., 2003, Walters et al., 2007, Clare and Bridge, 2005, Eickhoff et al., 2005, Bridge et al., 2005, Sigalovsky et al., 2006, Dick et al., 2012, Cohen-Adad et al., 2012, Lutti et al., 2013, Stüber et al., 2014). Using the ratio of T1- and T2-weighted (T1w and T2w) image intensities, Glasser and Van Essen (2011) detected the boundaries of myeloarchitectonically distinct cortical regions. Their method has subsequently been applied to link estimated cortical myelin content to cognitive performance and also to investigate the lifelong effect of age on cortical myelination (Grydeland et al., 2013). The myelin density map calculated from the ratio of T1w and T2w images has also been found to show significant correlation with the retinotopic map of the occipital cortex (Abdollahi et al., 2014).

The present paper builds on this prior work and explores a partial volume correction as applied to a large uniformly collected dataset. MR images common in biomedical research are subject to partial volume effect in which the intensity of the cortical gray matter (GM) can be contaminated by intensities of local white matter (WM) and cerebrospinal fluid (CSF). Here we introduce a partial volume correction algorithm to correct for this effect, and apply our technique to quantify the T1w/T2w intensity ratio as an estimate of myelin map (T1w/T2w myelin) of 1555 clinically normal 18 to 35 year old subjects using 1.2 mm isotropic voxel MRI.

It is important to note that T1w/T2w estimated myelin is not a pure measure of myelin but nonetheless an MR-accessible proxy. The ratio of the T1w and T2w images of a subject provides a unitless quantity that correlates with cortical myelination (Glasser and Van Essen, 2011; Glasser et al., 2013; Grydeland et al., 2013) but does not provide an absolute measure of myelin density. Also, iron, in addition to myelin, has been found to contribute to cortical MR image contrast. However, it has been shown that iron and myelin are often colocalized in the cortex (Fukunaga et al., 2010). Therefore, while we do not expect myelin to be the only factor contributing to the T1w/T2w image intensity ratio, a large fraction of the variation in T1w/T2w is likely due to variation in myelin density (more details in the MRI, myelin and comparison between subjects section).

With these caveats in mind, we (1) quantified partial volume effect on T1w/T2w myelin measurement, (2) used our technique to investigate the vertical distribution of myelin in the cortex, (3) quantified the effect of age on T1w/T2w myelin, and (4) investigated the relationship between cortical thickness, curvature and T1w/T2w myelin. We compared our results with high resolution in vivo MRI scans of five subjects and also a postmortem ex vivo brain scan to ensure that the difference in inner and outer layer myelination detected in the main dataset was not an artifact of mm-scale voxel size.