Multi-Seed Segmentation of the Primary Tumor Mass in Neuroblastoma using Opening-by-Reconstruction

R.H. Vu, R.M. Rangayyan, and G.S. Boag (Canada)

Keywords

Neuroblastoma, fuzzy connectivity, opening-by reconstruction, multi-seed segmentation.

Abstract

Segmentation of the primary tumor mass in neuroblastoma is an important problem. Segmentation could aid the ra diologist by facilitating reproducible, objective quantifica tion of the tumor’s tissue composition and volume. How ever, segmentation continues to be a difficult problem be cause the tumor is usually composed of inhomogeneous tissue types, some of which possess strong similarities in computed tomographic characteristics to contiguous non tumoral tissues. Previous work on the segmentation of the primary tumor mass focused on the removal of vari ous problematic tissues prior to segmentation using fuzzy connectivity. Although the results showed promise, fur ther work is required to minimize leakage of the result of segmentation. Opening-by-reconstruction is introduced in this work as a viable substitute for fuzzy connectivity be cause of its similarity to and computational advantages over the latter; furthermore, multi-seed segmentation can be in corporated into the procedure. The results suggest that opening-by-reconstruction is a more efficient segmentation method than fuzzy connectivity for the neuroblastic tumor, and is more suitable for the multi-seed approach.

Important Links:

Go Back