Abstract

Abstract INTRODUCTION Patients with cervical spinal cord injury (SCI) show different clinical outcomes. There is a significant association between the acute magnetic resonance (MR) imaging of cervical SCI and neurological recovery of cervical SCI. We speculated that principal component analysis (PCA), a dimension reduction procedure, would detect clinically predictive patterns in complex MR imaging and predict neurological improvements assessed by the American Spinal Injury Association Impairment Scale (AIS) and Japanese Orthopaedic Association (JOA) score. METHODS We performed a retrospective analysis of 50 patients with cervical SCI who underwent early surgical decompression less than 48 h after the trauma. We analyzed 7 types of MR imaging assessments: axial grade assessed by the Brain and Spinal Injury Center score (BASIC), longitudinal intramedurallry lesion length, spinal cord signal intensity on T1 and T2 weighted image, maximum canal compromise, maximum spinal cord compression, Subaxial Cervical Spine Injury Classification System. PCA was applied on these multivariate data to identify factors that contribute to recovery after cervical SCI following surgery. AIS conversion was evaluated at 6 mo. RESULTS Nonlinear principal component (PC) evaluation detected 2 features of MR imaging. PCA revealed PC 1 (40.6%) explaining the intramedullary signal abnormalities that were negatively associated with postoperative AIS conversion. PC2 (18.5%) suggested extrinsic morphological variables, but did not predict outcomes. The BASIC score revealed the significant overall predictive value for AIS conversion at six months (AUC 0.86). This result suggested that the intramedullary signal abnormalities reflect delayed neurological improvements even after early surgical decompressions in patients with cervical SCI. CONCLUSION PCA could be a useful data-mining tool to show the complex relationships between acute MR imaging findings in cervical SCI. This study emphasized the importance of multivariable intramedullary MR imaging as clinical outcome predictors.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call