System for Automatic Segmentation of Male Pelvis Structures from CT Images

描述(由申请人提供)：图像分割是一种常见的临床实践，用于从诸如计算机断层扫描和磁共振图像的体积图像中提取内部解剖对象的几何描述。当前实践的缺陷促使了一个重要的研究领域，涉及用于自动分割的统计可训练变形形状模型(DSM)。通常的方法是将数字空间模型应用于图像，并使模型经历一系列变形，以使模型和目标解剖对象(S)之间紧密匹配。在适当的统计框架中，通过数学优化来驱动变形。总体目标是开发和临床评估一个工作站，用于从CT图像中自动分割男性骨盆的解剖结构，用于放射治疗中的图像引导应用，使用基于一种特别强大的被称为m-reps的DSM的技术。分割的CT图像为放射治疗中的关键治疗计划和放射治疗决策提供指导。很可能为了这些目的而执行分割的次数比用于所有其他医疗应用的总和还要频繁。目前临床实践中的交互式等高线方法非常耗时和昂贵，并且等高线显示出显著的用户间和用户内部的变异性，这对依赖它们的临床决策产生了不利影响。建议的方法将克服这些缺点，并提高放射治疗癌症的有效性。

DESCRIPTION (provided by applicant): Image segmentation is a commonly performed clinical practice for extracting geometrical descriptions of internal anatomical objects from volume images such as computed tomography and magnetic resonance images. Shortcomings of current practice have motivated an important area of research involving statistically trainable deformable-shape models (DSMs) for automatic segmentation. The general approach is to apply a DSM to an image and cause the model to undergo a series of deformations converging to a close match between the model and the target anatomical object(s). The deformations are driven, in an appropriate statistical framework, by mathematical optimization. The overall aim is to develop and clinically evaluate a workstation for automatic segmentation of anatomical structures in the male pelvis from CT images for image- guided applications in radiation therapy using technology based on a particularly powerful class of DSMs called m-reps. Segmented CT images provide guidance for critical treatment planning and radiation delivery decisions in radiation therapy. It is likely that segmentation is performed more often for these purposes than for all the other medical applications combined. Current interactive contouring methods in clinical practice are extremely time consuming and expensive, and the contours demonstrate significant inter- and intra-user variabilities that adversely affect the clinical decisions that rely on them. The proposed methodology will overcome these shortcomings and improve the effectiveness of radiation therapy for treating cancer.