超高精度喉部切片图像数据集的建立

A high-precision digital images date set of larynx

目的:建立一个超高精度的喉部切片图像数据集。方法:将新鲜人喉标本以改良川本法进行包埋,采用LeicaCM3600XP冷冻切片机以10μm层厚进行连续切片,并用Nikon D750单反相机(2400万像素)对每个断面信息进行图像采集,分别以JPG和RAW格式储存。随后参照规则断面法,用Photoshop7软件裁剪所得图像进行尺寸校正和配准,以TIFF格式保存。结果:总计获得6 656张24位RGB原始图像,数据量为JPG格式97.5GB,RAW格式174GB,配准后图像最终分辨率为40μm/像素,数据总量107GB。结论:联合改良川本法、高分辨率单反相机和规则断面法,能够获得超高精度切片图像数据集,为类似于喉的人体局部解剖精细数据集的建立提供了基础。

Objective：To build a digital cross-sectional images date set of human normal larynx with high-precision. Method：A fresh frozen normal human larynx was embedded using Kawamot＇s Film Method and then serially sectioned at 10 μm with a LeicaCM3600XP Cryomacrotome, meanwhile digital images of the block surface were recorded after every section with a Nikon D750 camera and stored respectively as JPG and RAW files. All images were cropped to display a constant area on the block surface by photoshop7 and stored as TIFF files. Result.. A total of 6 656 images were recorded in 24-bit RGB resolution, the complete data file is 97.5 GB and 174 GB. After elimination of image size variation, all images were re-sampled at a final resolution of 40 μm/pixel and the complete data file is 107 GB. Conclusion：It＇s feasible to use the combination of Kawamot＇s Film Method, Digital Single Lens Reflex with high-resolution and regular block face-based alignment to build high-precision digital images date set, which could apply to some partial organs that resemble larynx.