基于水平集的前列腺磁共振图像分割方法研究

Research on the segmentation method of prostate magnetic resonance image based on level set

基于前列腺磁共振图像(MRI)特征信息及其病变好发特定区域等先验知识,针对前列腺内外轮廓全分割问题,提出基于边缘距离调整水平集演化(DRLSE)的前列腺MRI两步分割方法。在构建统一水平集能量函数的基础上,第1步基于前列腺MR的T1(纵向弛豫时间)图像实现其外轮廓分割,第2步在外轮廓约束限定条件下,基于前列腺MR的T2(横向弛豫时间)图像实现前列腺的内部轮廓分割,进而完成前列腺内外轮廓的全面有效分割。设计了前列腺分割的人机交互界面,对10个前列腺病例MR图像(含正常、增生和癌变共30幅)进行了分割实验研究,并采用Dice相似性系数(DSC)对分割结果进行评价分析,DSC值达到90%以上。实验结果表明,所提出的基于边缘DRLSE的前列腺MRI两步分割方法能够有效地实现前列腺内外轮廓的全面分割,非常接近于临床专家手动分割的理想结果,对前列腺疾病的临床诊断和治疗有较好的参考价值。

Based on some priori knowledge such as the feature information of prostate magnetic resonance images（MRIs） and its special zone different diseases often occur,aiming at the segmentation of prostate outer and inner contours,a two-step prostate MRI segmentation method based on edge distance regularized level set evolution（DRLSE） is proposed.On the basis of constructing a unified level set energy function,the first step is to realize the outer prostate contour segmentation based on the T1（longitudinal relaxation time） image of MR.The second step is to realize the inner prostate contour segmentation based on the T2（transverse relaxation time） image of MR under the condition of outer contour constraint; and then the effective whole segmentation of the prostate outer and inner contours is achieved.The human-machine interactive interface of the prostate segmentation was designed,and the image segmentation experiment study on ten prostate cases（including 30 images of normal,hyperplasia and cancerous prostates） was conducted.The Dice similarity coefficient（DSC） is used to evaluate and analyze the segmentation results,and the DSC value reaches to above 90%.The experiment results show that the proposed two-step prostate MRI segmentation method based on edge DRLSE can effectively realize the whole segmentation of prostate outer and inner contours,and the results are very close to the ideal ones obtained by clinic experts manually,which has good reference value for the clinical diagnosis and treatment of prostate diseases.