Neuroimaging scientists focusing on the human spinal cord.

To date, the only complete histological atlas of human cervical spinal cord was created by Bruce (1901).¹ This atlas contains high quality myelin and Nissl histology sections of every level human spinal cord segments created from a single sample. The assignment of distinct motor neuron clusters of the GM ventral horn, responsible to control specific limb and axial muscles, were made in the absence of neuromuscular connectivity data. Since then a number of studies have mapped the topography of the motorneuron clusters in relation to the limb muscles they control. While there are differences in the findings of these studies, they demonstrate a remarkable consistency in the pattern of forelimb and hindlimb muscle innervation in vertebrates.^2,3 Among the cervical motorneurons, these common features include the position of the phrenic nucleus at C4, the biceps and deltoid groups at C5 and C6, the forearm, triceps and pectoral groups at C7 and C8, and the manus muscle groups at C8 and T1.^3,4

The aim of this study is to create a high-resolution MRI atlas of the cervical spinal cord using ex-vivo samples. This atlas will be important to delineate grey and white matter structures that are normally invisible in in-vivo clinical imaging and that enables individual segments to be distinguished from one another with confidence.

Image acquisition: 13 ex-vivo cervical spinal cord specimens were scanned using 9.4T Bruker MRI. 0.2% Magnevist (Schering AG, Germany) was added to the specimens’ solutions for ~4 days prior to scanning. Glass tubes filled with Solvay Solexis Fomblin were utilised for imaging the specimen. High-resolution MRI data were acquired using a 3D fast low angle shot (FLASH) sequence with (TR/TE)=36/20ms, with an isotropic resolution of 80 μm with the scan time of 1.5 hours. Specimens were imaged with an overlap twice in order to obtain a full coverage within the homogenous RF coil and optimal gradients linearity.

Image processing and template: the start and end slices of each cervical segment were identified using the pattern of nerve roots from both ventral and dorsal sides, then the centre slice for each segments was identified accordingly using Osirix-generated 3D rendered volume (Figure1). The segment length and GM/WN areas were measured using Osirix. Spinal cord samples were classified into normal and pre-fixed types based on the positions of the rostral brachial motor neurons in the cervical levels. Finally, a high-resolution atlas of the normal-type samples was created using the protocol described in Figure 2.

From the examined spinal cords, 8 cords were classified as normal, and 5 cords were classified as pre-fixed. An average of T1 MRI, encompassing C3-C8 cervical segments, was created using four samples of normal-type cords (Figure 3). This average MRI template was compared with a histology cervical spinal cord atlas.¹ Comparison with the histology data shows the clarity of the visualisation of anatomical structures obtained by ex-vivo MRI. Human anatomy textbooks routinely state that the most rostral brachial motor neurons are found in the C5 segment.⁵ However, we have found that there are contributions of these rostral motor neurons founded in the C4 segment, at the same level as the main group of phrenic motor neurons. A comparison between the pre-fixed samples with Bruce 1901 histology atlas can be seen in (Figure 4).

Physical measurements of the lengths and areas were made to identify distinct variations between segments. It was found that the segment lengths and areas increased gradually from C1 to C6 and then declined steeply at C8 through the beginning of the thoracic segments (Figure5).

The assessment of ex-vivo human spinal cords shows significant variations between the segments. The variability in the position of human brachial motor neurons has been reported by many anatomical dissection studies. Approximately 21-48% of human population has prefixed cervical spinal cord ^6-10, where the rostral part of the brachial plexus arises from the C4 nerve instead of C5. In comparison, the normal-type cervical spinal cord has the brachial plexus arises from C5 to T1. Physical measurements show that as the cross section area in the spinal cord increases, the length of the segments and the areas of both gray and white matters increase accordingly. This information will be useful clinically to determine affected functions and aid in measuring atrophy associated with spinal cord injury.We have created a high-resolution MRI atlas of the cervical spinal cord, which will be useful as a template for registration and delineation of structures to improve diagnostic accuracy of clinical imaging.