3D Analysis of MR-imaged Articular Cartilage

描述(由申请人提供)： 骨关节炎和关节软骨损伤非常常见，通常会导致疼痛、肿胀和活动减少，导致功能障碍。患有这些疾病的患者的治疗成本是巨大的，仅在美国估计每年的经济成本就高达650亿美元。磁共振成像有助于在活体内对关节软骨进行无创成像。然而，人工分析是繁琐的、不准确的、不可重现的，并且没有商业上可用的计算机辅助工具来分析高分辨率3-D MR软骨图像数据。这项建议是为了从三维磁共振图像中对高度协调的关节(膝关节、髋关节、踝关节)的关节软骨进行定量分析的发展和可行性评估。第一阶段的提案将证明，软骨表面的识别(分割)可以高度自动化地完成，并且在关节接触区域最难分割的区域具有亚像素精度。该方法的主要创新之处在于，在3D中同时分割相邻骨骼和软骨的多个表面，而不是在2D中独立地逐边界和逐切片地分割，从而利用了3D上下文信息，并预期在3D分割性能上有所改善。 这一阶段-L建议的具体目的是： 1)提出了一种在体积高分辨率磁共振图像中同时检测软骨和骨表面的多表面检测方法。 2)开发了一种自动初始化这种多曲面分割的方法。 3)评估分割误差，并与现有独立标准进行比较，验证所开发的软骨分割和软骨形态评估方法的亚像素精度。 一旦关节接触区域的可行性被证明，未来的第二阶段将开发出全面分析整个关节的软骨形态和软骨组织特征的方法。因此，继续进行的第二阶段研究的目标将是建立和验证定量软骨分析的原型系统，该系统将被治疗骨关节炎的骨科临床医生、整形外科医生、肌肉骨骼放射科医生、流行病学和药理学研究人员以及磁共振扫描仪制造商使用。

DESCRIPTION (provided by applicant): Osteoarthritis and articular cartilage injuries are very common, typically causing pain, swelling and reduced mobility leading to functional impairment. The treatment cost for patients suffering from these conditions are massive with an estimated annual economic cost of $65 billion in the US alone. Magnetic resonance imaging facilitates non-invasive imaging of articular cartilage in vivo. However, manual analysis is tedious, inaccurate, and irreproducible and no computer-aided tool is commercially available to analyze high-resolution 3-D MR cartilage image data. This proposal is for development and feasibility assessment of quantitative analysis of articular cartilage in highly congruent joints (knee, hip, ankle) from 3-D magnetic resonance images. The Phase I proposal will demonstrate that identification (segmentation) of the cartilage surfaces can be accomplished highly automatically and with subvoxel accuracy in the most difficult-to-segment region of the joint contact area. The main innovation of the proposed approach is that the multiple surfaces of adjoining bones and cartilages will be segmented simultaneously in 3-D rather than independently border-by-border and slice-by-slice in 2-D, thus taking advantage of 3-D contextual information with an expected improvement in 3-D segmentation performance. The specific aims of this Phase-l proposal are to: 1) Develop a method for simultaneous multi-surface detection of cartilage and bone surfaces in volumetric high-resolution MR images. 2) Develop a method for automated initialization of such multi-surface segmentation. 3) Assess segmentation errors and demonstrate subvoxel accuracy of the developed methods for cartilage segmentation and cartilage morphology assessment in comparison with existing independent standards. Once feasibility is demonstrated in the joint contact region, the future Phase II will develop methods for comprehensive analysis of cartilage morphology and cartilage tissue characterization in the entire joint. Therefore, the goal of the continuing Phase II research will be to build and validate a prototype system for quantitative cartilage analysis that will be utilized by orthopedic clinicians treating osteoarthritis, orthopedic surgeons, musculoskeletal radiologists, and researchers for epidemiologic and pharmacologic studies, as well as by MR scanner manufacturers.