Spine Segmentation in CT volumes using geometric flows and shape priors

Each year more than 20,000 Canadians are treated surgically for lower back pain. Image guided surgical (IGS) techniques can reduce the number of complications that arise in about 20-30% of classical surgical procedures. Patients are often scanned for a CT pre-operatively and then, during the surgery, are scanned again using Ultrasound. The registration of the pre-operative CT scans with intra-operative ultrasound scans can help surgeons to perform surgeries with higher precisison. The segmentation of the spine in CT scans is, however, quite challenging. Inhomogenous regions inside and outside the vertebrae require the application of advanced segmentation methods. Signal loss at the boundaries of the vertebrae can further hinder the segmentation process. In this project we are working on developing precise and accurate models of the human vertebral column in collaboration with Dr. Louis Collins. To this end we are applying geometric flow methods combined with shape priors to automatically segment the entire spinal column as well as each vertebrae individually.

Surface Extraction on edge-enhanced flux-map image using Chan-Vese region based flow for lumbar region

Spine segmentation on edge-enhanced image using Chan-Vese region based flow for thoracic region

Axial, corononal and sagittal segmentations for the lumbar region

Shape prior based level set segmentation for the L4 vertebrae

Shape variations using PCA on the signed distance function. Red is the mean shape. The left gif shows the first ten negative deviations and the right shows the positive deviations.

Following are some open source demos created by Perklab using Slicer3D. The segmentations obtained by our method could be used in these settings.

Pedicle screw insertion in virtual reality

https://www.youtube.com/watch?v=F_UBoE4FaoY

How to segment multiple vertebrae in spine CT for 3D printing

https://www.youtube.com/watch?v=Uht6Fwtr9hE

Related Publications:
[1] A. Vasilevskiy and K. Siddiqi Flux Maximizing Geometric Flows. IEEE Transactions on Pattern Analysis & Machine Intelligence, 24(12), pp. 1565-1578, 2002.
[2] C.X.B. Yan, B. Goulet, J. Pelletier, D. Tampieri and D. L. Collins. Ultrasound-CT Registration of Vertebrae Without Reconstruction. International Journal of Computer Assisted Radiology and Surgery, 7(6), 901--909, 2012.