Osteosarcoma segmentation in CT images based on hybrid relative fuzzy connectedness

Osteosarcoma is very harmful and difficult to diagnose, which occurs most often in children and adolescents. Aiming at the difficulties in extracting tumor from Osteosarcoma CT images, which have few gray differences with the surrounding tissues, this paper proposes an interactive hybrid segmentation method combining modified relative fuzzy connectedness and confidence connected algorithm for Osteosarcoma CT images. In this paper, we select seed points from object and background areas interactively, and produce a rough segmentation from Osteosarcoma CT image by using the confidence connected method. The mean and variance values obtained from the rough regions are the initial values for fuzzy connectedness algorithm. In addition, we have developed a novel fuzzy spel affinity function, and compute fuzzy connectedness from each point to the seeds in object and background areas, then divide Osteosarcoma CT image into tumor tissue region and non-tumor tissue region by comparing the values of these two fuzzy connectedness. Finally, the segmented tumor tissue is rendered in 3D by the ray casting algorithm based on Graphics Processing Unit. According to the experiment results, the interactive hybrid segmentation method has a higher similarity index with the expert's manual segmentation compared to the original fuzzy connectedness algorithm. Also, it can segment Osteosarcoma tumor tissue more effectively and accurately, reducing the influence by the manual choices of the threshold and other parameters.