摘要
为实现前列腺超声图像的快速定位分割,提出了一种基于双向指数加权移动平均的算法来对分割过程中的曲线进行噪声滤除,从而保证分割精度和定位时间。首先,对经典的法向量轮廓边界算子进行改进使其结合邻域法向量信息提高分割精度,得到初分割曲线,该曲线上大部分点都已经位于真实边界点。然后,针对初分割后含噪声的边界曲线,使用双向指数加权移动平均算法对其进行两次去噪处理,得到最终的分割图像。实验结果表明,相比于热门的深度学习算法如U-Net、Mask R-CNN等,该方法有更优秀的分割效果,其DSC平均值为0.9582。由此可见,该方法在保留真实边界的同时,能够达到其他算法多次迭代的实验精度,在真实边界和噪声区域达到了良好的平衡。
Aiming at realizing the fast location and segmentation of prostate ultrasound image,an algorithm based on bidirectional exponentially weighted moving average is proposed to filter out the noise of the curve in the segmentation process,so as to ensure the segmentation accuracy and location time.First of all,the classical normal vector profile(NVP)is improved to improve the segmentation accuracy by combining the neighborhood normal vector information,and the initial segmentation curve is obtained,most of the points on the curve are already located at the true boundary points.Then,for the noisy boundary curve after the initial segmentation,the bidirectional exponentially weighted moving average algorithm is used to Denoise it twice to get the final segmented image.The experimental results show that this method has better segmentation effect than the popular deep learning algorithms such as U-Net,Mask R-CNN,and the average DSC of this method is 0.9582.Thus it can be seen that this method not only retains the real boundary,but also achieves the experimental accuracy of many iterations of other algorithms,and achieves a good balance between the true boundary and the noise area.
作者
李伟
LI Wei(Hubei Key Laboratory of Intelligent衍sion Based Monitoring for Hydroelectric Engineering,China Three Gorges University Yichang 43002,China)
出处
《长江信息通信》
2023年第4期43-47,共5页
Changjiang Information & Communications
基金
国家自然科学基金资助项目(No.61871258)。
关键词
前列腺分割
边缘检测
移动平均
平滑处理
图像去噪
prostate segmentation
edge detection
moving average
smoothing
image denoising
