• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>Congreso Argentino de Ciencias de la Informatica y Desarrollos de Investigacion >Development of a graphic interface for the three-dimensional semiautomatic glioblastoma segmentation based on magnetic resonance images
【24h】

Development of a graphic interface for the three-dimensional semiautomatic glioblastoma segmentation based on magnetic resonance images

机译：基于磁共振图像的三维半自动胶质母细胞瘤分割的图形界面的研制

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Glioblastoma is the most common and aggressive glioma in adults. Its complexity demands the development of methods able to maximize the capture of personalized information to let the design of patient-specific therapies, which can be achieved through radiomics studies. In this process, initial segmentation of the image is fundamental. In glioblastoma, the resulting segmented region of interest (ROI) must include the active tumor, its inner necrosis and the peripheral edema, a zone estimated to be infiltrated by tumor cells. In a first step, images corresponding to the different modalities of the MRI were registered to achieve spatial coincidence and the same three-dimensional resolution. In a second step the whole brain were segmented based on T1 images, to eliminate not-nervous tissues. Then the complete ROI region were determined through on a combination of FLAIR and T2 modalities and, finally, inner ROI components were defined working with the contrasted T1 modality. During these processes, K-means clustering, Chan-Vese active contours, adaptive thresholds, dilatation, erosion and replenishment algorithms were developed and grouped in the Matlab graphic interface RMIanalizer to interact with the user and visualize results. This interface can upload any format of medical image, segmentate semiautomatic and three-dimensionally the ROI components, and determine the estimated volume of each one. Preliminary results were compared with the "ground truth" cases submitted by the web database used, obtaining a Dice similarity coefficient of 0.886 +/- 0.0386 for the complete ROI region, over a total of 10 glioblastoma cases.
机译：胶质母细胞瘤是成人中最常见的和积极的神经胶质瘤。它的复杂性要求的能最大限度的个性化信息采集，让患者特异性治疗方法，它可以通过radiomics研究实现的设计方法的发展。在这个过程中，图像的初始分割是根本。在成胶质细胞瘤，关注区域（ROI）的所得到的分割区域必须包括活性肿瘤，其内坏死和外周性水肿，估计由肿瘤细胞浸润的区域。在第一步骤中，对应于MRI的不同模态的图像进行注册，以实现空间重合，并且在同一三维分辨率。在第二步骤中，全脑是基于T1的图像分割，以消除不神经组织。然后将完整的ROI区域是通过在FLAIR和T2模式和，最后，内ROI组分定义与对比T1工作模态的组合来确定。在这些过程中，K-means聚类，赞Vese主动轮廓，自适应阈值，扩张，侵蚀和补充算法进行了开发和在Matlab图形界面RMIanalizer分组，以与所述用户和可视化结果进行交互。此接口可以上传医学图像的任何格式，segmentate半自动和三维的ROI的部件，并确定估计的体积每一个。初步的结果与提出的使用的网络数据库，获得的0.886 +/- 0.0386一个骰子相似系数为完整ROI区域，在总共10成胶质细胞瘤情况下，“地面实况”的情况下进行了比较。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号