激光共聚焦序列图像基于特征的分割方法

A Segmentation Based on Feature for LSCM Sequence Images

采用fluo-4标记的乳鼠心肌细胞钙离子实时激光扫描共聚焦光学切片呈现为点状分布的荧光图像,并且受到噪声的严重干扰,细胞荧光图像在分布上没有连续的边缘;而采用处理静态荧光图像的方法处理这种图像存在很大的困难。该文就心肌细胞Ca2+离子定量研究过程中遇到的问题,提出一种基于马尔可夫场和空间点模式特征聚类相似性测度的自适应图像分割算法。该算法能以统一的标准从不同深度的光学切片中分割出需要分析的细胞钙离子活动区域,对其他利用激光共聚焦技术做动态分析的研究有很好的参考价值。

Region of interest （ROI） plays an important role in biomedical image analysis. The calcium ions laser scanning confocal microscope （LSCM） real-time optical sections of fluo-4 labeled rat ventricular myocyte show point distribution and are interfered by random noise. The fluorescent images have no continuous grayscale distribution and obvious edge. It is difficult to apply the methods which process static fluorescent images to our experiment. A self-adaptive segmentation is proposed based on spatial point pattern cluster features comparability in Markov field. The advantage of this method is that the ROI of calcium within all the disturbed optical sections can be properly segmented under a coincident standard. This method will benefit to other dynamic analysis researches using LSCM.