Averaging of diffusion tensor imaging direction-encoded color maps for localizing substantia nigra

Highlights

• A method for averaging diffusion direction-encoded color maps is presented.

• Application in a large sample of healthy controls shows improved signal-to-noise.

• Averaging in standard space helps distinguish substantia nigra from neighboring nuclei.

• Substantia nigra on average maps agrees with atlas locations and individual raters.

Abstract

Diffusion tensor imaging (DTI) is a form of MRI that has been used extensively to map in vivo the white matter architecture of the human brain. It is also used for mapping subcortical nuclei because of its general sensitivity to tissue orientation differences and effects of iron accumulation on the diffusion signal. While DTI provides excellent spatial resolution in individual subjects, a challenge is visualizing consistent patterns of diffusion orientation across subjects. Here we present a simple method for averaging direction-encoded color anisotropy maps in standard space, explore this technique for visualizing the substantia nigra (SN) in relation to other midbrain structures, and show with signal-to-noise analysis that averaging improves the direction-encoded color signature. SN is distinguished on averaged maps from neighboring structures, including red nucleus (RN) and cerebral crus, and is proximal to SN location from existing brain atlases and volume of interest (VOI) delineation on individual scans using two blinded raters.

Introduction

Diffusion tensor imaging (DTI) is a form of magnetic resonance imaging (MRI) that is sensitive to the diffusion of water molecules [1]. Since brain tissue is coherently organized and therefore presents barriers to random diffusion, DTI has proven extremely helpful for systematically mapping in vivo the white matter architecture of the human brain [2], [3]. DTI has also been used for mapping the location and subdivisions of subcortical nuclei, most notably the thalamus [4], which contains extensive connectivity with the cortical mantle. Other smaller nuclei, such as the subthalamic nucleus (STN) and substantia nigra (SN), can be visualized albeit with varying degrees of difficulty on individual subject DTI datasets [5], [6], but are nonetheless critical to locate. This is especially true for targeted electrical stimulation that can alleviate some symptoms of devastating neurological disorders like Parkinson’s disease [7], [8]. Yet, deep brain stimulation implants are complicated by failure rates during the targeting of small subcortical structures [9], [10]. Averaging of multiple direction-encoded color maps (DECM) is one simple method for potentially improving signal-to-noise and visualization of these structures within individual subjects for targeted surgical procedures.

Another challenge is visualizing consistent patterns of the orientation of the diffusion signal across groups of subjects. While the most widely referenced DTI brain atlases present color-coded diffusion orientation maps constructed from individual subject data [11], emerging techniques seek to combine data from multiple subjects in standard space to map consistent patterns of diffusion orientation [12]. These types of DTI population studies are important for building confidence about the average location and variability of brain structures that are difficult to localize. They are also important for tracking changes over time, as degeneration occurs slowly during aging and the accumulation of iron can affect the MRI signal [13], [14], [15].

The SN, a nucleus containing dopaminergic projection neurons, is vulnerable to degeneration even before the initial symptoms of Parkinson’s disease manifest [16]. The SN is difficult to resolve on a conventional T1-weighted MRI. However, several specialized MRI sequences and analysis techniques have been used to improve the ability to resolve the SN, including proton density-weighted MRI with short inversion-time recovery images [17], neuromelanin-sensitive MRI [18], quantitative T2 mapping [19], segmented inversion recovery ratio imaging [20], [21], [22], a combination of T2^⁎ and diffusion-weighted imaging [23], [24], connectivity-based parcellation using probabilistic diffusion tractography [25], and multispectral MRI sequences [26].

Several of these specialized techniques have been deployed with the goal of distinguishing the pars compacta (SNc) from the pars reticulata (SNr) subdivisions of the SN. However, these studies [17], [25] have provided ambiguous results regarding decreases in the volume of the SNc, the subdivision containing excitatory dopaminergic projections to the dorsal striatum, in Parkinson’s patients. While it has been demonstrated that the SN can be resolved on individual subject DTI DECM [6], [11], questions remain about the consistency with which the SN can be identified from these color maps across subjects. Additional questions remain about the location of SN on these color maps with respect to stereotactic coordinates of the SN referenced from available brain atlases.

To address these questions, we developed a simple method for averaging direction-encoded anisotropy-modulated color maps in standard space. Then we explored the improvement that signal averaging affords for localizing the SN in relation to other structures of the human midbrain. We chose the SN because its small size presents a continual challenge for localization, its critical landmark status for deep brain stimulation, and its importance as a candidate biomarker of early degeneration in patients at risk of developing Parkinson’s disease [27]. Our primary objective is to demonstrate, using a large sample of healthy controls, the utility of a simple and fast directional anisotropy averaging approach for visualizing the SN in relation to nearby structures, including the red nucleus (RN) and cerebral peduncle. Using two blinded raters, we then explore the location of the SN on average DECM with respect to the delineation of this structure on individual subject DTI datasets, and in relation to reported SN coordinates from the most frequently cited brain atlases.