Coagulated region recognition of microwave ablation based on texture analysis of B-mode ultrasound image

Microwave ablation (MA) is popular as a minimally invasive, painless and effective oncological treatment. The paper presents an approach of characterizing the thermal coagulation during microwave ablation using B-mode ultrasound image textures and pattern recognition technique. During the MA treatment, the series of B-mode ultrasound images were obtained. Textures of different ROIs such as gray level histogram and gray level co-occurrence matrices from the obtained ultrasound images were extracted separately. And then these textures were classified by support vectors machine (SVM). The experimental results have shown that the textures of coagulated and non-coagulated region have significant differences. The classification accuracy of SVM by 19 parameters is 0.9528. The method described in this paper is an effective way to realize thermal coagulation monitoring using texture features of B-mode ultrasound images.