Backscatter ultrasound physics for image segmentation and biological tissue characterization

背景：临床超声(US)系统包括用于结构成像的B型超声，用于血流特征的多普勒模式，以及用于组织变形和弹性评估的弹性成像模式。其他使用射频(RF)后向散射回波的方法也被开发用于许多器官的组织表征。框架和目的：现代定量超声(QUS)基于RF的组织表征方法主要依赖于两个主要策略。一种方法是对依赖于频率的反向散射系数(光谱分析)进行建模，以描述组织的微结构属性；而第二种方法是使用回波包络的一阶和二阶统计量来定义斑点属性。这些方法都还没有在临床扫描仪上实施。这项建议的框架包括统一这些不同的概念，以提出具有物理解释的诊断参数。这将通过考虑与频率相关的结构因子和回声统计的零差K分布的混合来实现。通过统一QUS的概念，我们的目标是识别具有特定组织微结构特征的互补参数，以达到病理组织分割和表征的目的。目标应用是动脉粥样硬化斑块、血栓、乳腺肿瘤和病变肌腱的成像。*新颖性和影响：US可以说是最难进行分割(和组织定征)的医学成像方式，因为图像对比度和结构清晰度比竞争对手的磁共振和计算机断层成像技术差得多。因此，大多数基于计算机的US分割方法需要解剖、几何、时间和/或图像物理先验。在这项资助中，我们在贝叶斯分割和组织表征模型的框架内提出了新的图像物理概念，该模型可以包括解剖、几何和时间先验来动态跟踪图像序列中的器官。这个项目在技术上具有挑战性，以基础声学物理为基础，应该会直接影响人类健康，因为目标应用广泛，并覆盖我的实验室活跃的临床研究领域(临床数据的获取得到了其他正在进行的拨款的支持)。*结论：US成像是目前放射学和医学物理领域中单位数量增长最快的，因为它是非侵入性的，手持，与其他成像技术相比相对便宜，还因为它可以用于床的大小。根据使用的应用程序和方式，US也是敏感和具体的。这项赠款计划中提出的基于QUS的方法的发展应该会影响这一势头，并为临床医生提供新的成像方式，以便更好地进行诊断和治疗。

BACKGROUND: Clinical ultrasound (US) systems include B-mode for structure imaging, Doppler modes for flow characterization, and elastography modes for tissue deformation and elasticity assessments. Other methods using radiofrequency (RF) backscatter echoes have also been developed for tissue characterization of numerous organs.*******FRAMEWORK and OBJECTIVE: Modern quantitative US (QUS) RF-based methods for tissue characterization mainly rely on two main strategies. One approach consists in modeling the frequency-dependent backscatter coefficient (spectroscopy analysis) to describe microstructural properties of tissues; whereas a second approach is to use 1st and 2nd order statistics of the echo envelope to define speckle properties. None of these approaches has been implemented yet on clinical scanners. The framework of this proposal consists in unifying these different concepts to propose diagnostic parameters with a physical interpretation. This will be done by considering a frequency dependent structure factor and mixtures of homodyned K-distributions of echo statistics. By unifying QUS concepts, we aim identifying complementary parameters with specific signatures of tissue microstructures for the purpose of pathological tissue segmentation and characterization. Targeted applications are imaging of atherosclerotic plaques, blood clots, breast tumours and diseased tendons.*******NOVELTY and IMPACT: US is arguably the hardest medical imaging modality upon which to perform segmentation (and tissue characterization) as image contrast and structure definition are much worse than competing magnetic resonance and computed tomography imaging technologies. Accordingly, most computer-based methods for US segmentation require anatomical, geometrical, temporal and/or image physics priors. In this grant, we propose novel image physics concepts in the framework of a Bayesian segmentation and tissue characterization model that can include anatomical, geometrical and temporal priors to dynamically track organs within an image sequence. This project is technically challenging and based on fundamental acoustic physics, and should directly impact human health as targeted applications are broad and cover active clinical areas of research of my laboratory (clinical data acquisitions are supported by other active grants).*******CONCLUSION: US imaging represents today the largest grow in term of number of units in the radiology and medical physics fields because it is non-invasive, hand-held, relatively inexpensive compared to other imaging technologies, and also because it can be used on the bed size. Depending on the application and modality used, US is also sensitive and specific. The development of QUS-based methods, as proposed in this grant program, should impact this momentum and provide clinicians with new imaging modalities for better diagnostic and therapy follow-up.***