Abstract
Brain-tumor segmentation is an important clinical requirement for brain-tumor diagnosis and radiotherapy planning. It is well-known that the number of clusters is one of the most important parameters for automatic segmentation. However, it is difficult to define owing to the high diversity in appearance of tumor tissue among different patients and the ambiguous boundaries of lesions. In this study, a nonparametric mixture of Dirichlet process (MDP) model is applied to segment the tumor images, and the MDP segmentation can be performed without the initialization of the number of clusters. Because the classical MDP segmentation cannot be applied for real-time diagnosis, a new nonparametric segmentation algorithm combined with anisotropic diffusion and a Markov random field (MRF) smooth constraint is proposed in this study. Besides the segmentation of single modal brain-tumor images, we developed the algorithm to segment multimodal brain-tumor images by the magnetic resonance (MR) multimodal features and obtain the active tumor and edema in the same time. The proposed algorithm is evaluated using 32 multimodal MR glioma image sequences, and the segmentation results are compared with other approaches. The accuracy and computation time of our algorithm demonstrates very impressive performance and has a great potential for practical real-time clinical use.
Highlights
Magnetic resonance (MR) imaging technology has been widely applied to medical diagnosis systems, and the accuracy of many diagnosis systems is mainly based on the quality of the images acquired
Because the computing speed and convergence of the classical mixture of Dirichlet process (MDP) method is not very good for clinical image clustering, we introduce anisotropic diffusion and Markov random fields (MRF), which are combined with the classical MDP models to construct our algorithm
A clinical high-grade glioma T1C MR image is segmented by the MDP/MRF algorithm
Summary
Magnetic resonance (MR) imaging technology has been widely applied to medical diagnosis systems, and the accuracy of many diagnosis systems is mainly based on the quality of the images acquired. The images obtained by magnetic resonance imaging usually contain heavy noise and the effects of the biasing field, which will degrade the quality of the images and make the subsequent postprocessing of the images, such as segmentation, classification, and detection, difficult. The noise and biasing-field effects in MR images sometimes even affect the evaluation of human segmentation. Segmentation for MR images automatically becomes a very challenging task. The essential segmentation task is to label the different parts of the object image. The class number of the parts for the segmented image is a very important parameter. As is well-known, the number of clusters for medical image segmentation is difficult to initialize as a constant before clustering. Most classical segmentation methods for medical images [1, 2] specify the number of clusters before clustering as a matter of clinical experience, even some advanced methods estimate the number of clusters from educated guesses or prior knowledge [1, 2]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
