The Duke University Cervical Spine MRI Segmentation Dataset (CSpineSeg).

This work describes a publicly available dataset, the Duke University Cervical Spine MRI Segmentation Dataset (CSpineSeg), consisting of 1,255 cervical spine magnetic resonance imaging (MRI) examinations from 1,232 patients collected from the Duke University Health System. CSpineSeg also includes expert manual semantic segmentations of vertebral bodies and intervertebral discs for 481 patients. This dataset aims to provide a resource for training and evaluation of deep learning segmentation models and facilitate cervical spine research. Along with the dataset, we present a deep learning segmentation model which could be used as a benchmark in cervical spine segmentation tasks. Our segmentation model achieves a Dice Coefficient of 0.916, demonstrating the feasibility of utilizing CSpineSeg to train segmentation models.