Application of color image segmentation in the diagnosis of bladder cancer

The diagnosis of bladder tumors mainly relies upon cystoscopic examination. Some researchers used an in vivo staining technique to improve diagnosis. However, this method heavily depends upon the interpreter's experience, and is subject to errors. In this paper, we used methylene blue in vivo staining in combination with image analysis to improve the accuracy of diagnosis of bladder tumors. We developed an approach to segment the bladder tumor from the image. There are two stages in the segmentation process. First, the probabilistic relaxation using contextual information is employed to reduce the local ambiguities and obtain the initial classification of the objects. Second, in order to resolve the chromatic properties, the bayesian algorithm is used to further classify objects of different colors. Various color models were used in the segmentation process and the results were compared to obtain an optimal color model. The results showed that the proposed approach can segment the bladder tumor and identify the tumor boundary successfully. This technique could be an auxiliary tool for doctors in the future.