Research
Spine Segmentation
3D Segmentation of spinal vertebra is a crucial step in studying spinal related disease. Complexity of vertebra shapes as well as noise and incomplete image formation are a major source of difficulty. Research undertaken has presented a successful framework for accurate segmentation, capable of dealing with a great deal of noise in an image. The techniques used include edge-mounted Willmore flow, and a prior shape kernel density estimation, which are used to guide the level set segmentation framework.
Accuracy of the algorithm is found to be 89.32±1.70%, using the DICE similarity coefficient. Inter- and intra-observer variation agreements are 92.11% and 94.94% respectively.
Citations: [·]