Fast image segmentation for pulmonary lesions using hybrid level set model

The lung cancer radiotherapy treatment widely depends on adequate diagnosis. The radiologists intend to reach an image segmentation efficiency in terms of accuracy and low computation cost. However, the pulmonary lesions segmentation is still considered as a challenging task due to the noise and intensity inhomogeneity present in Computed Tomography (CT). In this study, we proposed to accelerate the nonlinear adaptive level set model, using the Bayesian rule, by incorporated the double well potential in the regularization term to get accurate and fast pulmonary lesion segmentation in CT images. We have tested the proposed method on different sized and localized lesions. All the images were taken from the database without any preprocessing. The experimental results show significant speed improvement without losing the precision of segmentation.