An Effective Organ Extraction Framework for Serialized Visible Human Slices Based on Spectra Analysis and Skeleton Graffiti

摘要

Extracting the regions of interest (ROIs) from human body slice images in the Visible Human Project (VHP) data set is a crucial task in VHP research. In this paper, an effective organ extraction framework for serialized Visible Human slices is proposed. This framework mainly contains two parts: first, a spectral analysis strategy is utilized to finely extract the organ regions; second, a double skeleton graffiti method is employed to automatically extract indispensable geometric input information for the next slice image. Using this framework, color slice images of VHP can be automatically and serially processed. As a result, the primary organs in VHP body slices can be extracted clearly and accurately. In addition, satisfactory reconstructed 3D models of the obtained organs can be designed. Such 3D organ models could be useful for many related applications (such as virtual surgery). This extraction framework also provides new insights for processing similar sequential image sets.