Magnetic resonance neurography (MRN) is a technique commonly used for evaluating abnormal conditions of peripheral nerves [1]. One of the recently developed MRN techniques is the 3D nerve-SHeath signal increased with INKed rest-tissue RARE Imaging or 3D SHINKEI [2]. Our work aims at evaluating 1.5T SHINKEI images to establish the conditions of normalcy in the brachial plexus and then to differentiate normal anatomy from the ones having avulsion, lesions or nerve compressions by using the second order Gray Level Co-occurrence Matrix (GLCM)[3]. The goal of our work is to develop an image analysis methodology that correlates with the clinician’s conventional classification and nerve conduction studies (NCS) for the brachial plexus injury.

The SHINKEI technique combines an iMSDE preparation pulse set for flow signal suppression and fat-suppression prep-pulse SPAIR that suppresses the fat signal followed by T2 weighted 3D-TSE acquisition. Our study was approved by the institutional ethical committee and this ongoing retrospective study comprises data sets from 14 volunteers and 3 patients with root avulsion condition acquired at 1.5T scanner (Achieva, Philips Healthcare) using TR of 2500 ms, effective TE of 127 ms, voxel-size of 1.2 x 1.2 and iMSDE prepulse for 50ms.

Nifti images of SHINKEI based MRN were analyzed using an in-house developed MATLAB based image processing script. Prior to the image analysis, the radiologist selected the ROI as depicted in Figure 1. During analysis, we assumed, the value P (i, j | d (theta) ) is the relative frequency with which two pixels with intensities i and j are separated by a pixel distance d in the direction theta that occurs within a given neighborhood The number of rows and columns of the GLCM is equal to the number of gray levels in the image ROI as each entry (i,j) in GLCM means the number of times that the pixel with value i occurred horizontally (theta = 0) adjacent to a pixel with value j. We have studied the GLCM in the 0, 45, 90 and 135 degree directions from pixels within the ROI. The data was arranged group-wise for each directional contrast, with each row representing each volunteer. Similarly, the maximum of the other three directions is compared with contrast in 45 degrees.

Out of the fourteen volunteers imaged in the study, the image regions were separated into roots, trunk and cords parts of the brachial plexus. The contrast in 45 degrees was tested for difference from the maximum contrast among other three directions in the volunteers’ root, trunk and cords, using paired t test and significance was tested (Table 1).

We have observed that the contrast provided by the GLCM was high in the perpendicular direction of the nerve direction originating from the spine outwards in a normal condition while in a diseased state, the contrast was more homogeneous. The mean ratio of contrast_45 to the maximum contrast in 0, 90 and 135 degree directions 1.83±0.05 in root of normal volunteers while it is 1.1±0.08 in arbitrary unit (a.u.) for patients with nerve root avulsion (Figure 2).

We have demonstrated our method based on contrast property from the second order gray level Co-occurrence Matrix to classify at the level of roots, trunk and cords. In addition, our preliminary results demonstrate that the he normal brachial plexus root can be differentiated from the pathological conditions by the value of contrast obtained from the GLCM. Our ongoing work includes analysis of higher SNR SHINKEI based MRN data obtained from 3T scanner and further analyzes the efficacy of our newly developed image analysis methodology.