The invention discloses a method and a device for searching an optimal
spinal canal path centerline. Imaging scans, such as
computed tomography (CT) scans, are used to diagnose rib and spinal fractures. The spinal image is typically reformatted with straightened CPR (curved planar reformatted). A
dynamic programming-based technique is disclosed to find an optimal path from
spinal canal segmentation. The disclosed methods and devices overcome
spinal canal segmentation errors that lead to spinal canal gaps and unrealistic image distortions. The disclosed techniques are insensitive to segmentation errors even if there is high segmentation
noise. The disclosed techniques generate a binary three-dimensional image corresponding to a scanned spinal image, perform a density
estimation to create a three-dimensional having the same size as the original binary three-dimensional image, and perform a
dynamic programming-based optimal spinal canal path lookup using an optimal path lookup strategy from thick to thin to accelerate the optimal spinal canal path lookup.