人正常离体子宫薄层切片数据集的自动配准

摘要

目的:通过改良及优化后的配准方法实现人正常离体子宫薄层切片数据集的自动配准.方法:在Photoshop中处理原始断层图像,获取定位杆图像和标本图像(子宫、双附件及部分阴道)；在Matlab中计算定位杆图像中四个定位杆的质心坐标值,将其作为定位杆坐标值,取所有定位杆图像定位杆坐标值的平均值作为基准坐标；每张定位杆图像的定位杆坐标与基准坐标对比,获取每张定位杆图像的投影变换参数并根据该参数对相对应的标本定位杆图像及定位杆图像进行投影变换；将校正后的定位杆图像再在Photoshop中处理,获取只包含第二个定位杆的图像,同样在Matlab中计算其坐标值,并基于其坐标值将配准后的标本定位杆图像裁剪成大小一致图像；选取配准后两张大小相差较大的相邻图片,在Photoshop中将配准前的两张相邻图片及配准后的两张相邻图片分别进行重叠,调节其中一张图片的透明度,将合成的图片保存.结果:配准前两张相邻图片中的标本大小差距较明显,而配准后两张相邻图片中的标本差距很小.通过该配准方法,成功地实现了人正常离体子宫薄层切片数据集的自动配准.结论:基于Photoshop软件和Matlab软件的图像配准方法具有配准精度高,运算量小,易于编程实现等优点,通过改良及优化后,可用于人正常离体子宫薄层切片数据集的配准.%Objective: Using the registration method after reforming and optimization to automatically register the human normal in-vitro uterus thin-layer data collection. Methods: Firstly, the primary images were import into Photoshop software, and then the locating rod images and the specimen-locating rod images (uterus, bilateral accessories and part of vagina) were obtained. Secondly, we calculated the centroid coordinate value of the four locating rod in the locating rod images as the locating rod coordinate values, and the average coordinate value of the locating rod of all the locating rod images as the reference coordinate value by using Matlab software. Thirdly, we obtained the projective transformation parameters of every locating rod image by comparing the locating rod coordinate value of every locating image with the reference value, and got the specimen-locating rod images and locating rod images according to the parameters. Fourthly, The corrected locating rod images were then import into Photoshop software, and then the images only containing the second locating rod were got. We calculated the coordinate value of the first locating rod in Matlabrnsoftware and cropped the specimen—locating rod images into the same size based on the coordinate value. Finally, we selected twornadjacent images of equal size, overlapped the two images before registration and after registration in Photoshop software, regulatedrnthe transparency of one image, and then saved the synthesized image. Results: The size difference between the specimen in therntwo sdjacent images before registration is very large, where is very small after registration. The automatic registration of the humanrnnormal in—vitro uterus thin—layer data collection was realized successfully by using this registration method. Conclusion: The imagernregistration method based on Photoshop software and Matlab software has some advantages, such as high registration accuracy, lessrncomputation, easy to implement, and can be used for the registration of the human normal in-vitro uterus thin-layer data collectionrnafter reforming and optimization.