东篱科研大数据发现系统（DRDS）

位置：成果数据库 > 期刊 > 期刊详情页

软袋组合盖质量视觉检测系统研究

ISSN号：1002-8331
期刊名称：计算机工程与应用
时间：2013
页码：-
分类：R323.1[医药卫生—人体解剖和组织胚胎学;医药卫生—基础医学]
作者机构：[1]湖南大学,电气与信息工程学院,湖南长沙410082, [2]中南大学,湘雅医学院,附属湖南省肿瘤医院,放疗科放疗中心和PET/CT中心,湖南长沙410013
相关基金：国家自然科学基金（61271382,61301254,61471166）;湖南省肿瘤医院科研平台建设基金
相关项目：基于视觉的药液产品质量在线检测方法及关键技术研究

关键词：正电子发射断层扫描, 核磁共振成像, 图像分割, 自适应区域生长, 靶区勾画, 头颈癌, positron emisson computerized tomography, magnetic resonance imaging, image segmentation, adaptive regiongrowing, segmentation of target volume, head and neck cancer

中文摘要：

目的：为了有效解决单独使用正电子发射断层扫描（PET）和核磁共振成像（Mm）影像勾画大体肿瘤靶区（G墨V）存在的肿瘤、水肿及其周围正常组织区分难题。方法：首先选取PET图像上包含肿瘤区域的感兴趣区域（ROI）中标准摄取值（suv）最大的体素点作为肿瘤区域生长算法的初始种子点，在PET和潮影像上分别进行第一阶段自适应区域生长。然后从其勾画的肿瘤PET靶区内自动获取肿瘤的最小SUV值，并联合肿瘤MRI靶区自适应区域生长的最佳阈值构建第二阶段肿瘤PET和MRI联合区域生长准则，进行第二阶段区域生长，完成PET与MRI融合靶区勾画。结果：与单独使用PET和单独使用MRI影像的自适应区域生长分割结果相比，参考两位经验丰富的临床放疗专家手工勾画的鼻咽癌MRI GTV，本文方法勾画的融合GTV与MRIGTV具有最高相似性，且同时具有较高灵敏性和较高特异性。结论：本文方法可实现头颈部肿瘤PET与MRJ融合大体肿瘤靶区自适应高精度勾画。

英文摘要：

Objective To effectively distinguish tumors from edemas and the surrounding normal tissues in the positron emission computerized tomography （PET） image or magnetic resonance imaging （MR/） image for segmenting gross target volume （GTV）. Methods In the volume of interest containing tumor region in PET image, the voxel with the maximum standard uptake value （SUV） was chosen as the initial seed of the adaptive region growing （ARG） algorithm. The first stage ARG was respectively applied on PET images and MRI images. The minimum SUV in the segmented target volume based on PET images was automatically acquired and combined with the best threshold value of the tumor ARG based on MR/to determine the growth criterion of the second stage ARG based on both PET and MR/images. The second stage ARG was carried out to complete the segmentation of combined target volume based on PET and MRI images. Results Compared with the segmentation results of ARG by independently using PET images or MR/images, the combined GTV segmented by the proposed method had the highest similarity with the GTV in the MR/ image of nasopharyngeal carcinomas segmented by two experienced radiation oncologists, and achieved higher sensitivity and specificity. Conclusion The proposed method achieves adaptive high precision segmentation for the GTV of head and neck cancer based on PET combined with MR/.

同期刊论文项目