Variational principle for ultracoustic artifact correction and signal segmentation

摘要

We present a variational framework, titled BLA (Backscatter-Levelset-Attenuation joint estimation model), to simultaneously correct regional attenuation artifact and segment anatomical structures in ultrasound images. The BLA is posed as a mathematical inverse problem and solved via functional minimization. The echoes are modeled using acoustic backscatter and attenuation of structures. A weak-form model is constructed by incorporating the regularizations on the backscatter, attenuation, and structure boundaries with level set boundary representations. The BLA needs not to solve partial differential equations over irregular image domains, thus boosting the computational efficiency toward optimizing ultrasound imaging and signal segmentation.