基于非线性形变模型的双向凝胶图像校正

发布时间：2018-08-15 16:29

【摘要】：在生物化学和生命科学等领域,随着现代科学水平的进步,蛋白质组学已经成为前沿的研究领域。在蛋白质的鉴定及生理病理学中的蛋白质功能测定中,首先利用双向凝胶电泳技术对复杂的蛋白质组织进行分离,扫描存档,完成双向凝胶电泳图像的采集。凝胶图像中含有大量蛋白点,需要利用计算机软件进行分析。分析同一样本在不同时期凝胶图像中对应蛋白点对的变化,从而提取出感兴趣蛋白点对进行研究。在获取凝胶图像的过程中,由于蛋白点的移动变形、外部环境的变化、仪器的多样性和仪器产生的噪声等问题可能造成同一样本的凝胶之间出现图像失真。为了更好地进行蛋白质组的分析,必须在凝胶图像的预处理阶段消除失真,完成图像的校正,以便更好地实现凝胶图像的匹配。凝胶图像的校正,就是消除几何失真对图像产生的影响。本论文以凝胶图像的几何失真校正为研究重点,基于多种形变方法,完成凝胶图像的校正。主要研究内容如下:(1)本论文研究了基于薄板样条函数(Thin Plate Spline,TPS)的形变,通过手动标记控制点,找到一个通过部分控制点的弯曲能量最小的光滑曲面。实验结果表明,对于凝胶图像的非线性几何失真,校正效果比较明显,但手动选点会存在一定的误差,而标记点的位置偏移,会导致图像的校正偏差;(2)基于B样条函数的形变,同样是非线性形变方法的一种,用自由形变模型的网格点作为图像的控制点,根据控制点局部控制性和计算效率考虑,构建三次均匀B样条形变模型。通过改变网格间距的大小,网格控制点周围的点随之改变,其校正效果也有明显的变化,必须根据失真程度选择合适的网格大小。仿真结果表明,较TPS形变校正来讲,校正效果有所改善。但在失真程度比较严重、网格间距较大时,部分控制点形变效果欠佳;(3)由于凝胶图像失真的中心拉伸特点,“笑脸”形变模型可以对图像进行整体的形变,消除这种微笑失真。本论文将“笑脸”形变模型和B样条形变模型有效的融合在一起,提出一种新的校正算法,充分利用“笑脸”全局形变和B样条函数局部形变的特点。仿真实验结果表明,这种方法和其他形变校正方法相比,形变校正效果更佳。刚性变换、仿射变换等线性变化,算法简单,运算速度快,但是只适合图像的平移、旋转等线性几何失真的校正,对非线性的几何失真效果甚微。本论文将几种典型的非线性形变方法用在凝胶图像的失真校正上,对比分析校正效果得出,本论文所提出的形变校正方法,能将失真图像的细节信息和先验知识进行有效结合,在保证形变平滑性的同时又能得到良好的校正效果,并具有较高的计算效率,适合实际研究中使用。
[Abstract]:With the progress of modern science, proteomics has become a frontier research field in biochemistry and life sciences. In the identification of proteins and the determination of protein function in physiology and pathology, two-dimensional gel electrophoresis was used to separate the complex protein tissues, scan and archive, and complete the collection of two-dimensional gel electrophoresis images. Gel images contain a large number of protein spots, which need to be analyzed by computer software. The changes of protein point pairs corresponding to the same sample in different periods of gel images were analyzed, and the interested protein point pairs were extracted for study. In the process of obtaining gel images, the distortion of the gel images of the same sample may occur due to the movement and deformation of the protein points, the change of the external environment, the diversity of the instruments and the noise generated by the instruments. In order to better analyze the proteome, the distortion must be eliminated in the pre-processing stage of the gel image, and the correction of the image should be completed, so that the matching of the gel image can be realized better. The correction of gel image is to eliminate the effect of geometric distortion on the image. In this paper, we focus on geometric distortion correction of gel image, and complete the correction of gel image based on various deformation methods. The main research contents are as follows: (1) the deformation based on thin plate spline function (Thin Plate splines is studied in this paper. By manually marking control points, a smooth surface with minimum bending energy through partial control points is found. The experimental results show that the correction effect is obvious for the nonlinear geometric distortion of the gel image, but there will be some errors in the manual selection of the point, and the deviation of the position of the mark point will lead to the correction deviation of the image. (2) based on the deformation of B-spline function, It is also one of the nonlinear deformation methods. The mesh points of the free deformation model are used as the control points of the image, and the cubic uniform B-spline deformation model is constructed according to the local control and computational efficiency of the control points. By changing the size of the mesh spacing, the points around the grid control point will change, and the correction effect will also change obviously. The appropriate mesh size must be selected according to the distortion degree. The simulation results show that the correction effect is better than that of TPS deformation correction. However, when the distortion degree is serious and the mesh spacing is large, the deformation effect of some control points is not good. (3) because of the central stretching characteristics of gel image distortion, the "smiling face" deformation model can deform the image as a whole and eliminate the smile distortion. In this paper, the "smiling face" deformation model and the B-spline deformation model are effectively fused together, and a new correction algorithm is proposed to make full use of the global deformation of "smiling face" and the local deformation of B-spline function. The simulation results show that this method is more effective than other deformation correction methods. The linear changes such as rigid transformation and affine transformation are simple and fast, but they are only suitable for the correction of linear geometric distortion, such as translation and rotation, and have little effect on nonlinear geometric distortion. In this paper, several typical nonlinear deformation methods are applied to the distortion correction of gel images. By comparing and analyzing the correction results, it is concluded that the distortion correction method proposed in this paper can effectively combine the details of distorted images with prior knowledge. It can ensure the smoothness of deformation at the same time get a good correction effect, and has a higher calculation efficiency, which is suitable for practical research.
【学位授予单位】：山东师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】