Medical ultrasound image segmentation and biological tissue characterization

医学超声图像分割和生物组织表征

• 批准号: 138570-2011
• 负责人: Cloutier, Guy
• 金额: $ 5.03万
• 依托单位: Université de Montréal
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2015
• 资助国家: 加拿大
• 起止时间: 2015-01-01 至 2016-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=569364
• 关键词: Medical; ultrasound; image; segmentation; biological

Ultrasound imaging has traditionally been used to diagnose medical conditions and is also developed for therapy purpose where the precise identification of the targeted organ or structure is a prerequisite. Ultrasound image segmentation consists in delineating boundaries of physical structures with the objective of localizing them, measuring their dimensions, characterizing the underlying tissue properties or to support inverse problem formulation for more advance characterizations. Traditional and more recent image segmentation methods applied to ultrasound images rely typically on pixel-based, region-based, intensity gradient-based, texture-based, local-phase-based and statistical speckle property-based features integrated into active contour, deformable template, watershed, neural network, Bayesian, multi-agent, level set and fast-marching methods. The major challenges in segmenting ultrasound images versus other radiology imaging modalities such as MRI, CT and angiography are the noisy nature of those images, missing information due to shadowing and signal attenuation, low contrast, and movements associated with dynamic image acquisitions and ultrasound image-guided interventions. Consequently, most methods require anatomical, geometric, intensity, temporal and image physics priors. A promising innovative avenue is to incorporate additional image physics priors to address even more challenging problems, such as the sub-segmentation of structures with common cellular properties within pre-segmented organs. In addition to speckle statistics, tissue characterization features based on the spectral content of radio-frequency echoes can be integrated into the nonlinear optimization function of segmentation methods to enhance performance. This will be the objective of this research program that has been synergically designed to support my ultrasound research activities on carotid and coronary atherosclerotic plaque characterization, breast cancer diagnosis and imaging of blood clots. The proposed program will also support artivities of my laboratory on diseased tendon characterization, endocardiac border detection and photo-acoustic imaging.

超声成像传统上用于诊断医疗状况，并且也被开发用于治疗目的，其中精确识别目标器官或结构是先决条件。超声图像分割在于描绘物理结构的边界，目的是定位它们、测量它们的尺寸、表征底层组织特性或支持反演问题表述以实现更高级的表征。应用于超声图像的传统和最新的图像分割方法通常依赖于基于像素、基于区域、基于强度梯度、基于纹理、基于局部相位和基于统计散斑属性的特征，这些特征集成到活动轮廓、可变形模板、分水岭、神经网络、贝叶斯、多智能体、水平集和快速前进方法中。与 MRI、CT 和血管造影等其他放射成像方式相比，分割超声图像的主要挑战是这些图像的噪声性质、由于阴影和信号衰减而丢失信息、低对比度以及与动态图像采集和超声图像引导干预相关的运动。因此，大多数方法需要解剖、几何、强度、时间和图像物理先验。一个有前途的创新途径是结合额外的图像物理学先验来解决更具挑战性的问题，例如在预先分割的器官内对具有共同细胞特性的结构进行细分。除了散斑统计之外，基于射频回波频谱内容的组织表征特征可以集成到分割方法的非线性优化函数中以增强性能。这将是该研究项目的目标，该项目旨在支持我在颈动脉和冠状动脉粥样硬化斑块特征、乳腺癌诊断和血栓成像方面的超声研究活动。拟议的计划还将支持我的实验室在患病肌腱特征、心内膜边界检测和光声成像方面的活动。