An efficient and semiautomatic segmentation method for 3D surface reconstruction of the lumbar spine from Magnetic Resonance Imaging (MRI)

A new semiautomatic technique for the segmentation and reconstruction of 3D vertebral anatomy from MRI is presented where five cadaveric human lumbar spines and one ovine spine were used. The MR images were subjected to contrast enhancement, anisotropic diffusion filtering, and thresholding was selected as the preferable segmentation technique to create 3D surfaces from the MRI datasets. The reconstructions were manually cleaned using commercial software. The resulting reconstructed surface included discrete vertebral bodies with distinct separation between the spinous processes. CT reconstructions were used to assess the accuracy of the reconstructed spinal anatomy from MRI