自适应滤波结合改进T-Snake模型的甲状腺超声图自动分割方法

It proposed a new segmentation method of thyroid ultrasonic image based on T-Snake model. Firstly, a window-based anisotropic diffusion filter, combined with an adaptive weighted median filter, effectively eliminated speckle noise in thyroid ultrasonic images. Secondly, the method, based on the traditional T-Snake model and the adaptive region energy and expansion force, extracted the discontinuous boundary and weak boundary effectively, and realized the automatic segmentation of thyroid ultrasonic image. Finally, it set the model parameters and experimented with the clinical data. The results showed that the average relative difference is less than 5% and the average relative overlap is more than 91%. It verifies the feasibility of the proposed method.