Introducing Willmore Flow into Level Set Segmentation of Spinal Vertebrae

Poay Hoon’s recent work, related to her PhD, has been accepted for publication in IEEE Transactions on Biomedical Engineering. Congratulations!

Introducing Willmore Flow into Level Set Segmentation of Spinal Vertebrae

by Poay Hoon Lim, Ulas Bagci, and Bai Li


Segmentation of spinal vertebrae is a crucial step in studying spinal related disease or disorders. However, limited work has been done on 3D segmentation of the spinal vertebrae, especially with ground truth validation. The complexity of vertebrae shapes, with gaps in the cortical bone and boundaries,as well as the noisy, incomplete or missing information from the images have undoubtedly increased the challenge for image analysis. In this paper, we introduce edge-mounted Willmore flow to our level set segmentation framework that integrates prior shape and local geometrical information for 3D segmentation of spinal vertebrae. While the shape energy draws the level set function towards a range of possible prior shapes, the edge-mounted Willmore energy captures the localized geometry information and smooths the surface during the level set evolution. Evaluation of the 3D segmentation results from CT images of spinal vertebrae with ground truth validation demonstrate the effectiveness of our approach. Evaluation based on Dice similarity coefficient and Hausdorff distance showed that our results achieved an overall accuracy of 89.32 +-  1.70% and 14.03  +- 1.40mm respectively. Whereas the inter and intra-observer variation agreements are 92.11+-1.97%, 94.94+-1.69%, 3.32+-0.46mm and 3.80+-0.56mm on our dataset.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s