Whiplash is the colloquial term for neck injuries caused by sudden extension of the cervical spine. Patients with chronic whiplash associated disorder (WAD) can experience neck pain for many years after the original injury. Researchers have found some evidence to suggest that chronic whiplash is related to the amount of intra-muscular fat in the cervical spine muscles. Hence, an important step towards developing a treatment for chronic WAD is a technique to accurately and efficiently measure the amount of intra-muscular fat in the muscles of the cervical spine. Our proposed technique for making this measurement is to automatically segment the cervical spine muscles using a fused volume created from multi-modal MRI volumes of the cervical spine. Multiple modes are required to enhance the boundaries between the different muscles to assist the following automatic segmentation process. However, before these multiple modes can be fused it is first necessary to accurately register these volumes. Hence, in this paper, we have proposed a new non-rigid multi-modal registration algorithm using the sum of conditional variance (SCV) with partial volume interpolation (PVI) similarity measure and Gauss-Newton (GN) optimization for the accurate registration of multi-modal cervical spine MRI volumes. The performance of the proposed approach is compared with the existing SCV based registration algorithm and the sum of the conditional squared deviation from the mode (SCSDM) method. The experimental results demonstrate that the proposed approach provides superior performance than the best existing approaches.