3217

Assessment of the Specificity and Sensitivity of DTI Metrics for Evaluation and Diagnosis in Degenerative Cervical Myelopathy
Guangqi Li1, Xiaodong Ma1, JinChao Wang2, Donghang Li2, Xiao Han2, Wen Jiang3, Xiaoguang Cheng3, and Hua Guo1

1Center for Biomedical Imaging Research, Department of Biomedical Engineering, School of Medicine, Tsinghua University, Beijing, China, 2Department of Spine Surgery, Beijing Jishuitan Hospital, Beijing, China, 3Department of Radiology, Beijing Jishuitan Hospital, Beijing, China

Synopsis

Diffusion Tensor Imaging (DTI) can detect diffusion information of water molecules, and is used to diagnose the severity of degenerative cervical myelopathy (DCM). However, the diagnostic capability of DTI metrics is not fully investigated. In this study, DTI metrics are employed to evaluate the spinal cord function in preoperative DCM patients and healthy volunteers. Nonparametric t-test results show that MD, FA and RD have significant differences between patients and healthy volunteers. In addition, ROC results indicate that FA has higher sensitivity, RD has higher specificity for evaluation and diagnosis in DCM.

Introduction

Degenerative cervical myelopathy (DCM) is one of the major causes of spinal cord dysfunction. Diffusion Tensor Imaging (DTI) parameters can describe diffusion of water molecules at a microcosmic scale. It has been used for clinical evaluation of DCM. Previous studies show that fractional anisotropy (FA) significantly reduces at stenotic levels in DCM patients [1-5]. However, the accuracy of using DTI metrics for evaluation and diagnosis in DCM is not fully validated. In this study, we sought to investigate the specificity and sensitivity of DTI metrics for evaluation and diagnosis in DCM.

Methods and materials

Image acquisition
All MR images were acquired on a Philips Ingenia 3.0T (Philips, Best, The Netherlands) with a 16-channel head-neck coil. 57 preoperative DCM patients (40 Males and 17 Females, age 44-75 years, mean 60.57±6.71 years) were recruited under IRB approval in Beijing Jishuitan Hospital as the case group. Diffusion data with two b values (1000 s/mm2 and 2000 s/mm2) and 32 diffusion directions were acquired using a single-shot EPI sequence with a SENSE factor of 2. A reduced FOV (160×60mm2) and saturation bands were applied to decrease the distortion. In-plane resolution=1.5×1.5 mm2 and TE/TR=77/4470ms. 17 transverse slices with a thickness of 4mm and the gap of 2mm were used to cover cervical levels from C2 to C7. Also, sagittal T2-weighted (T2W) and transversal T2*-weighted (T2*W) anatomical images were obtained. 21 healthy volunteers (15 Males and 6 Females, age 52-65 years, mean 56.86±5.08 years) were scanned with the same imaging protocol as the control group.
Image processing
First, spinal cord toolbox (SCT) [6] was used to correct for potential motion in the diffusion images. Then DTI parameters (MD, AD, FA and RD) were calculated using FSL [7]. ROI analysis was performed at the most compressed level for the patients and at all cervical levels for the healthy volunteers using DTI-studio [8]. Specifically, ROIs were manually drawn on all parameter maps with the axial T2*W image (Fig. 1) as a reference. The mean value of each parameter within the ROI was recorded.
Statistical analysis
The preoperative patients were divided into three subgroups according to the location of the most compressed level (C3/4, C4/5 and C5/6). The sample sizes were 12, 19 and 23, respectively. Three preoperative patients with most compressed level of C6/7 were excluded due to a small sample size. Considering the sample size and the actual distribution of the DTI metrics in three subgroups, nonparametric independent Mann-Whitney U test was selected to test the differences of each parameter between the DCM patients and the healthy volunteers. In addition, Receiver Operating Characteristic (ROC) analysis was employed to investigate the specificity and sensitivity of DTI metrics for clinical assessment of DCM. Furthermore, Youden Index (YI) of the ROC curve was calculated to obtain the cut-off value for each parameter. All statistical analysis was performed in SPSS with version 23. P<0.05 was considered statistically significant for all tests (*P<0.05, **P<0.005, ***P<0.0005, and ****P<0.0001).

Results and Discussion

The t-test results for the DTI parameters are shown in Fig. 2. MD, FA, and RD have significant differences between the DCM patients and healthy volunteers at three cervical levels (C3/4, C4/5 and C5/6). The FA value of the patients is remarkably less than that of healthy volunteers (P<0.0001), which is consistent with previous studies [1-5]. Moreover, MD and RD values from the patients are higher due to axonal degeneration and demyelination [9]. These parameters can reflect the microstructural differences of spinal cord between patients and healthy volunteers, indicating that they have the robust capability to assess pathological information of the spinal cord in DCM. As shown in Fig. 3, all AUCs of MD, FA and RD are larger than 0.90 at the three different stenosis levels. The specificity and sensitivity of DTI parameters for evaluation of DCM are listed in Table 1. Statistical results demonstrate that MD, FA and RD can be used in diagnosis of DCM. FA has better sensitivity and RD has better specificity for evaluation and diagnosis in DCM.

Conclusion

In this study, the specificity and sensitivity of DTI metrics for evaluation and diagnosis in DCM is investigated. Statistical results reveal that MD, FA and RD can capture the differences of spinal cord microstructural change between patients and healthy volunteers with high accuracy. Specifically, when used to evaluate the spinal cord dysfunction in DCM, FA has higher sensitivity and RD has higher specificity.

Acknowledgements

No acknowledgement found.

References

[1] Rindler, Rima S., et al. "Spinal Diffusion Tensor Imaging in Evaluation of Preoperative and Postoperative Severity of Cervical Spondylotic Myelopathy: Systematic Review of Literature." World neurosurgery 99 (2017): 150-158.

[2] Guan, Li, et al. "High-resolution diffusion tensor imaging in cervical spondylotic myelopathy: a preliminary follow-up study." NMR in Biomedicine 30.10 (2017).

[3] Jones, J. G. A., et al. "Diffusion tensor imaging correlates with the clinical assessment of disease severity in cervical spondylotic myelopathy and predicts outcome following surgery." American journal of neuroradiology 34.2 (2013): 471-478.

[4] Guan, Xiaofei, et al. "Diffusion tensor imaging studies of cervical spondylotic myelopathy: a systemic review and meta-analysis." PloS one 10.2 (2015): e0117707.

[5] Facon, David, et al. "MR diffusion tensor imaging and fiber tracking in spinal cord compression." American journal of neuroradiology 26.6 (2005): 1587-1594.

[6] De Leener, B., Lévy, S., Dupont, S. M., Fonov, V. S., Stikov, N., Louis Collins, D., Cohen-Adad, J. (2016). SCT: Spinal Cord Toolbox, an open-source software for processing spinal cord MRI data. NeuroImage. http://doi.org/10.1016/j.neuroimage.2016.10.009

[7] M. Jenkinson et al., “FSL.” Neuroimage 62(2), 782–790 (2012)

[8] Jiang, Hangyi, et al. "DtiStudio: resource program for diffusion tensor computation and fiber bundle tracking." Computer methods and programs in biomedicine 81.2 (2006): 106-116.

[9] Ellingson, Benjamin M., Noriko Salamon, and Langston T. Holly. "Advances in MR imaging for cervical spondylotic myelopathy."European Spine Journal 24.2 (2015): 197-208.

Figures

Fig. 1 Upper row: T2*W images and DTI parameter maps (MD, AD, FA and RD) at a stenotic level from a DCM patient; Lower row: T2*W images and DTI maps at a normal level from a healthy subject. It is shown that there is an obvious deformation of spinal cord in the DCM patient.

Fig. 2 The t-test results for comparison of DTI metrics between the preoperative patients and the healthy volunteers. MD, FA and RD were significantly different between the patients and the healthy volunteers at three compressed levels (C3/4, C4/5 and C5/6). The FA value is remarkably lower, MD and RD value are much higher in the DCM patients. P<0.05 was considered statistically significant for all tests (*P<0.05, **P<0.005, ***P<0.0005, and ****P<0.0001). No significant differences were marked with N.S.

Fig. 3 The ROC curve of DTI metrics for evaluating the spinal cord function in DCM patients and healthy volunteers. From top to bottom, the results of MD, AD, FA and RD at three different compressed levels were shown. Except for the AD parameter, the AUCs of MD, FA and RD all exceed 0.90 at C3/4, C4/5 and C5/6 levels.

Table 1. The ROC analysis for DTI parameters.
1 Preoperative DCM patients were divided into three subgroups according to their most stenotic level.
2 P<0.05 was considered statistically significant for all tests (*P<0.05, **P<0.005, ***P<0.0005, and ****P<0.0001). No significant differences were marked with N.S.
3 Cut-off point is confirmed by the maximum Youden Index (YI), and the corresponding specificity and sensitivity were listed in rightmost two columns.

Proc. Intl. Soc. Mag. Reson. Med. 26 (2018)
3217