摘要
计算机辅助诊断通过对乳腺磁共振成像(MRI)中肿块区域的自动分割和测量为医生提供定量的诊断依据。对分割过程中不同阶段的多种算法进行实验对比,以此探索更具稳定性和准确性的分割方案:空间模糊C均值(s FCM)聚类算法在肿块的初始定位中具有抗噪声能力和稳定性强的优点,而GVF snake模型在精细分割中对局部轮廓具有较好的收敛性;结合两种算法,并运用MRI序列帧间灰度分布相似、肿块位置/形状相近的原理,最终提高整个序列的分割精度与稳定性。
CAD（ computer-aided diagnosis） could be applied to assist the doctors in the diagnosis of breast cancer, by providing quantitative parameters of breast tumors with the automatic segmentation and measurements of the tumor regions in breast MRI（ magnetic resonance imaging） slices. In order to find out a stable and accurate segmen-tation scheme,a variety of segmentation algorithm of different stages had been carried upon the contrast experiments. sFCM（ spatial Fuzzy c-means clustering algorithm） was applied to locate the tumor roughly for its high denosing abil-ity and stability. And then GVF snake model was utilized to segment the tumor accurately for its high convergence of local boundary. Finally,relevant theory of inter-frame images was used to improve the segmentation accuracy of the whole MRI sequence,since the gray distribution and the positions of the tumors are always very similar in the adja-cent slices.
出处
《传感技术学报》
CAS
CSCD
北大核心
2015年第3期387-395,共9页
Chinese Journal of Sensors and Actuators
基金
国家自然科学基金项目(60705016
61001215
61205200)
浙江省自然科学基金项目(LY12F03003)
关键词
医学图像处理
乳腺肿块分割
帧间相关性
核磁共振成像
模糊C均值
SNAKE模型
medical image processing
tumor segmentation of breast masses
inter-frame correlation
magnetic reso-nance Imaging
Fuzzy c-means clustering algorithm
snake model
