构建MRI统计形变模型的子宫肌瘤超声图像分割方法

摘要

超声图像分割是实现高强度超声聚焦治疗中术中引导的重要步骤,而超声图像的低信噪比使得分割的结果往往不精确,易造成手术中聚焦区域有偏差.为此提出一种基于MRI先验形变模型的超声图像分割方法.首先分别从MRI图像集和超声图像集中训练基于特定病人的统计形变模型和初始化分割模型；然后利用超声图像中子宫肌瘤的部分显著性轮廓对分割模型进行初始化；最后使用统计形变模型对超声图像的分割结果进行约束,并对结果进行优化得到最终的结果.与其他分割方法进行比较的实验结果表明,该方法能够有效地提高高强度超声聚焦治疗的精度和效率.%Segmentation of ultrasonic images is an important step in guide of the intra-operative high intensity focused ultrasound therapy.However,the low signal-to-noise ratio of ultrasonic image makes segmentation result imprecise,which causes the deviation of focus area during surgery.Therefore,an ultrasonic image segmentation method based on MRI prior deformation model is proposed.Firstly,the statistical deformation model and initial segmentation model based on patient-specific are trained respectively from MRI set and ultrasonic image set; then segmentation model is initialized by the partial salient profile of fibroid; finally,the result of segmentation is constrained and optimized by statistical deformation model.Comparing with another segmentation methods,the final result of experiment shows that this method can effectively improve the accuracy and efficiency of the high intensity focus ultrasound therapy.