中文摘要：

微流控芯片可实现单细胞分析,而对单个细胞分析,能够掌握更准确更全面的细胞信息,可以克服以往群体分析中平均结果对个别信息掩盖的局限性,对疾病的早期预防和诊断具有重要的科学意义。本文根据早期癌症细胞通过微流控芯片中的弯道时变形与正常细胞不同的理论,采用Grabcut和Snake相融合的单细胞图像分割算法来精确定位和提取单细胞轮廓,实现单细胞的形变分析。首先,本文在图像分割之前引入Perona-Malik模型,增强图像边缘的同时减弱噪声,使定位更加准确。其次,利用Canny和Snake模型获得Grabcut初始化矩形框。最后,通过Grabcut算法实时精确地提取单细胞轮廓。实验结果表明：本文算法结合了Snake算法和Grabcut算法的优点,在无人工交互的条件下,细胞轮廓平均正确分割率达到93.7%,能够满足医学单细胞分析的要求。

英文摘要：

The single-cell analysis, achieved on the microfluidic chip, is able to grasp more accurate and comprehensive cells information, has important scientific significance for the prevention and early disease diagnosis. Based on the theories that the deformability of the early cancer cells are different from normal cells when they go around a curve in the mierofluidie chip. So a new cell segmentation algorithm, which fuses classical grabcut algorithm with snake algorithm, is proposed to accurately extract a single cell outline, and achieve the deformation analysis of a single cell. Firstly, the morphological operators are used to eliminate small bright spot noise. Secondly, canny and snake algorithm are utilized to get the position of one cell, which can be regarded as the initialization of the grabcut algorithm. Finally, accurate cell contour are extracted by grabeut algorithm. Experimental results show that the proposed algorithm combines the advantages of snake algorithm and grahcut algorithm. In the absence of human interaction, the average correct segmentation rate is up to 93.7%, can meet the medical requirements of single cell analysis.