Imaging Appearance of Magnetic Resonance Myelography in Normal Population: Employing Three-dimensional Sampling Perfection with Application Optimized Contrasts Using Different Flip-angle Evolutions (3D-SPACE) Sequence

Characterizing the normal distribution of cerebrospinal fluid (CSF) in the spine is crucial for an accurate assessment of CSF leakage and other abnormalities. Magnetic resonance myelography (MRM) is a noninvasive diagnostic method that is commonly used to evaluate the spinal distribution of CSF. Our aim was to evaluate the anatomical distribution of CSF in the spine of healthy individuals, by MRM using the three-dimensional sampling perfection with amplification-optimized contrasts using flip-angle evolutions (3D-SPACE) sequence. Twenty-one healthy volunteers underwent whole-spine MRM imaging using the 3D-SPACE sequence. MRM images were reconstructed with 5-mm axial multiplanar reconstruction (MPR) and maximum intensity projection (MIP) at each spine level. Two radiologists evaluated CSF distribution from the spinal canal in the MPR and MIP images, using 7-point (types A–G) and 3-point (grades 0–3) classification systems, respectively. Inter-reader agreement was calculated with the kappa coefficient (κ). Reader 1/reader 2 evaluated 46/53, 67/57, 4/6, 0/0, 0/0, 7/5, and 2/5 cervical spine (C-spine)-level MPR images corresponding to types A–G, respectively (κ = 0.74). Numbers at the thoracic spine (T-spine) level were 185/186, 41/44, 8/5, 0/0, 0/0, 16/11, and 2/6, respectively (κ = 0.69), and at the lumbar spine (L-spine) level were 46/58, 25/23, 9/5, 0/0, 0/0, 19/13, and 6/6, respectively (κ = 0.50). Inter-reader agreement for MPR images at the whole-spine level was considered good (κ = 0.69). Reader 1/reader 2 evaluated 0/0, 13/13, 8/8, and 0/0 MIP images at the C-spine level corresponding to grades 0–3, respectively (κ = 0.80). Numbers at the T-spine level were 10/14, 8/5, 3/2, and 0/0, respectively (κ = 0.67), and at the L-spine level were 5/2, 10/8, 6/11, and 0/0, respectively (κ = 0.26). Inter-reader agreement for MIP images at the whole-spine level was considered good (κ = 0.61). In conclusion, T2-weighted MRM with 3D-SPACE sequence imaging can be a useful technique to detect the normal distribution of CSF in the spinal canal. Understanding the normal distribution of CSF in the spinal canal is necessary to achieve an accurate diagnosis of CSF leakage. Correspondence Author to: Hung-Chieh Chen Department of Radiology, Taichung Veterans General Hospital, Taichung, Taiwan No.1650, Sec. 4, Taiwan Boulevard, Taichung 407, Taiwan J Radiol Sci 2014; 39: 1-6 MR myelography in normal population 2 J Radiol Sci March 2014 Vol.39 No.1 Imaging modalities that have been used to detect CSF leakage in the spine include computed tomographic (CT) myelography, radioisotope cisternography, and magnetic resonance myelography (MRM). As a radiation-free and noninvasive imaging method, MRM is the most frequently used of these approaches to diagnose CSF leakage. Commonly reported MRM imaging findings indicative of CSF leakage include a triangular-shaped expansion of the neural sleeve, an irregular linear signal lateral to the neural sleeve, distended spinal epidural veins, and extradural fluid collection [1, 2]. Axial multiplanar reconstruction (MPR) images are used in conjunction with maximum intensity projection (MIP) images for an accurate determination of the CSF leakage site. However, MRM imaging has some disadvantages. In particular, there is considerable inter-reader variability in the interpretation of MRM images obtained using different pulse sequences. Furthermore, the normal population may also present MRM images showing abnormalities mimicking CSF leakage. This ambiguity is a potential source of confusion. Here, we aimed to determine whether MRM with a newly developed pulse sequence, three-dimensional sampling perfection with application optimized contrasts using different flip-angle evolutions (3D-SPACE), could be used for effectively detecting the CSF distribution in the spinal canal of normal individuals. To our knowledge, this is the first study examining the spinal distribution of CSF in a normal population using MRM with the 3D-SPACE sequence. MATERIALS AND METHODS This study was approved by the Institutional Review Board (IRB) of Taichung Veterans General Hospital. Subjects Twenty-one healthy adults (10 males, 11 females; age range, 20–50 years; mean age, 28.0 ± 9.7 years) without any history of headaches or known spinal disease were recruited. Magnetic resonance myelography Subjects underwent MRM imaging of the whole spine using the 3D-SPACE sequence. The parameters used for MRM were as follows: TR of 3000 ms, TE of 560 ms, fat suppression, isotropic voxel size of 0.9 mm, matrix size of 320 × 320, and field of view (FOV) of 200 mm. The generalized autocalibrating partially parallel acquisitions (GRAPPA) imaging reconstruction technique with an acceleration factor of 2 was used with the 3D-SPACE sequence. 3D-SPACE sequence images were acquired in the coronal plane. Axial whole spine MPR images were obtained as 5-mm slices and MIP images were reconstructed at the cervical, thoracic, and lumbar spine levels. Image evaluation To evaluate the CSF distribution and spinal canal morphology, the MRM images were reconstructed by MPR and MIP processing. We used the modified MIP image grading scale proposed by Yoo et al. [4], which we considered to be more applicable to a normal population (Fig. 1). Additionally, we created a classification system to categorize MPR images (Fig. 2), based on our experiences to interpret spinal magnetic resonance (MR) images. To our knowledge, there are no reports characterizing the MR imaging profile of the spine in a normal population. We also did not find any studies examining the imaging profile of CSF distribution in a normal population using CT myelography, which could be used as a reference. Two radiologists (with 4 and 9 years of experience, respectively) independently evaluated the MRM images for each level of the spine. The image quality and contrast were evaluated by 3-point scale, as follows: grade 3 – minimal noise and aboveaverage image contrast; grade 2 – presence of noise, but sufficiently clear for a diagnosis to be made, and average image contrast; and grade 1 – presence of significant noise levels that impaired image interpretation and poor image contrast. Statistical analysis The results of two readers were analyzed and the interreader agreement between two readers was evaluated with the kappa coefficient (κ).