An automated region of interest (ROI) segmentation tool on diffusion tensor imaging of the cervical spinal cord

Abstract

Objective: Cervical myelopathy (CM) is a chronic spinal degenerative disease that is induced by cervical spinal compression by degenerative vertebrae and disc. Diffusion Tensor Imaging (DTI) was recently studied to evaluate the pathological changes of cervical spinal cord under the affect of CM. In prior studies, researchers applied region of interest (ROI) technique to segment and analyze the DTI images in corresponding area, which is a labor intensive process and may involve biases from operators. We proposed and validated a knowledge-based automated segmentation algorithm that enables automated ROI definition on images of the DTI in cervical spinal cord in this study. Methods: Thirty-two control subjects and eighteen patients diagnosed with CM were recruited to undergo routine Magnetic Resonance (MR) sequence and DTI for validating the knowledge-based and self-developed segmentation algorithm. Intra-observer and inter-observer reliability on the developed algorithm were evaluated. Result: The study demonstrated the developed algorithm was fast and reliable in performing column-specific ROI definition in the control and uninvolved levels in the CM groups that enabled the possibility to conduct large scale research in health population and detect the CM involved spine levels. However, it requires further optimization to enhance the successful rate of interaction with the compressed spinal cord.