摘要
目的为建造具有东方人种特征的中国数字化虚拟人,获取中国女性人体数字图像的数据集。方法通过广州市遗体捐献中心,选取一位因食物中毒急性死亡19岁女性尸体,经CT、MRI采集图像后,在股动脉灌注红色填充剂,冷冻包埋,在JZ1500A立式铣床上,以400 mm刀盘切削,转速850 r/m,断面间距0.2 mm,用第三代富士数码相机摄像。结果按中国解剖学会的《中国人解剖学数据》加以评价,VCH-F1体型指数为94%,有较大的代表性。共获取断面图像8 556个,图像总像数649万,每个文件数据量为17.5 MB,VCH-F1数据集总量149.7GB。共编制6个不同版本,分别为VCH-F1(149.73GB、6.60GB、0.27GB)及VCH-F1头部数据集(1.93GB、0.67GB、0.46GB)。结论借鉴国外经验和教训,VCH-F1在下列几个方面有所改进和提高:(1)尸体选材代表性较大;(2)血管标记鲜明,对后续图像配准和分割提供了良好条件;(3)立式切削较卧式切削更符合人体生理姿势;(4)本文采用的人体的冷存,一次装夹,双温冰库,大口径刀盘处理,能提高图像质量和工作效率。VCH-F1的实验数据集,在若干关键技术上有所改进和提高,构建了切削精度高,血管显示鲜明,体型代表性良好,为我国今后虚拟人体的物理化和生理化,提供了一个高质量的框架。
Objective To establish digitized virtual Chinese No.1 female (VCH-F1) image database. Methods A 19 years old female cadaver was scanned by CT, MRI, and perfused with red filling material through formal artery before freezing and em- bedding. The whole body was cut by JZ1500A vertical milling machine with a 0.2 mm inter-spacing. All the images was pro-duced by Fuji FinePix S2 Pro camera. Results The body index of VCH-F1 was 94%. We cut 8 556 sections of the whole body, and each image was 17.5 MB in size and the whole database reached 149.7 GB. We have totally 6 versions of the database for different applications. Conclusions Compared with other databases,VCH-F1 has good representation of the Chinese body shape, colorful filling material in blood vessels providing enough information for future registeration and segmentation. Verti-cal embedding and cutting helpd to ratain normal human physiological posture, and the image quality and operation efficiency were improved by using various techniques suth as one-time freezing and fixation, double-temperature icehouse, large-diame-ter milling disc and whole body cutting.
出处
《第一军医大学学报》
CSCD
北大核心
2003年第3期196-200,209,共6页
Journal of First Military Medical University
基金
Supported by National"863"Development Project in High-Tech Research (No 2001AA231031 and No 2002AA231021)
by Guangdon Provin-cial Science and Technology grant (2002B30611)
关键词
虚拟人
血管铸形
立式切削
立式包埋
Virtual Human
blood vessels casting
vertical cutting
vertical embedding
