Segmentation of Spinal Cord (SC) part has a major role in assessing spinal cord atrophy. Spinal Cord segmentation is not similar to that of brain segmentation, although Magnetic Resonance Imaging (MRI) sequence are greatly deployed for spinal cord examination. There is always a great challenge in spinal cord MRI segmentation which has been investigated by many researches. Also, considerable accuracy and degree of complexity for segmentation have been presented and elucidated in prevailing researches. A new approach namely combining Iterative Random-Walk (RW) solver and a Cuckoo Search Algorithm (CSA) has been suggested, thus facilitating direct homogenous and heterogeneous SC measurements comparison. An interactive RW solver with CSA is greatly utilized for complete cascaded pipelining in automatic manner. The initialization of automatic segmentation pipeline is done through powerful voxelwise classification via Online Kernel Learning (OKL) classifier. Therefore, SC topology refinement is done iteratively along with cascading of RW-CSA solvers for attaining proper segmentation outcomes in less iteration, even for cases including bone fractures and lesions. The segmentation experimental outcome mainly relies on MRI images indicating achievement of improved accuracy when compared with prevailing approaches.