Lung-specific ultrasound beamforming for diagnostic imaging

项目摘要 准确诊断和监测肺部疾病，包括新冠肺炎引起的迫切需要， 通过超声成像广泛解决。诊断和监测肺部疾病的标准模式有 X射线成像和计算机断层扫描（CT）由于其广泛的诊断能力。超声波可能不会 通常被认为是主要的肺部成像模态，然而在专家用户手中， 相对于CT，敏感性和特异性范围为90%至100%。 对于非专家用户，肺部超声图像的解释可能是复杂的，因为超声不能 穿透软组织/空气界面。因此，肺部超声依赖于对成像“伪影”的解释。 这些声音似乎来自肺部的空气空间深处，但实际上是来自肺部的复杂回响。 胸膜界面这些反射携带关于潜在的肺部病理学的信息。这种间接成像 临床解释方法与软组织成像有根本的不同， 直接来自被成像的结构。然而，目前使用的延迟求和波束成形方法 在超声系统中，对于肺成像和软组织成像是相同的。缺乏对 在复杂的软组织/空气界面处的基本声学仍然是合理设计 超声成像序列可以直接与肺部声学相关，并且对疾病更敏感。 为了克服这一挑战，我们建议开发和验证新的超声成像和波束形成 使用基于物理学的方法建立超声成像和 肺部的疾病状态。我们假设设计的超声波束形成技术 专门针对肺部及其复杂的混响物理学将生成更高质量的图像， 临床可解释性和诊断能力。我们将开发声学模拟工具， 人体和肺部疾病的实验校准，以准确地代表相关的 混响物理学，例如A线和B线伪影。空间相干波束形成器，它依赖于 混响作为对比度和机器学习波束形成器的来源将被设计和优化， 检测肺部疾病。这些波束形成器将在可编程扫描仪上实现，并与 常规B型成像。如果成功，该提案将产生更多的超声成像方法， 对肺部疾病敏感，具有更清晰的临床可解释性，可用于当前的超声成像 系统.

PROJECT SUMMARY Accurate diagnosis and monitoring of lung disease, including the urgent need arising from Covid-19, could be widely addressed by ultrasound imaging. The standard modalities that diagnose and monitor lung disease are X-ray imaging and computed tomography (CT) due to their extensive diagnostic capabilities. Ultrasound may not be normally thought of as a primary lung imaging modality, however in the hands of an expert user it has a sensitivity and specificity ranging from 90% to 100% relative to CT. For non-expert users the interpretation of lung ultrasound images can be complex because ultrasound cannot penetrate the soft-tissue/air interface. Thus, lung ultrasound relies on the interpretation of imaging "artefacts" that appear to come from deep inside the air space of the lung, but are actually complex reverberations from the pleural interface. These reflections carry information about the underlying lung pathology. This indirect imaging and clinical interpretation approach is fundamentally different from imaging in soft tissue, where echos come directly from the structures being imaged. Nevertheless, delay-and-sum beamforming methods currently used in ultrasound systems are identical for lung imaging and soft tissue imaging. The lack of understanding of the fundamental acoustics at the complex soft-tissue/air interface remains an impediment to the rational design of ultrasound imaging sequences that can relate directly to lung acoustics and would be more sensitive to disease. To overcome this challenge, we propose to develop and validate new ultrasound imaging and beamforming methods using a physics-based approach that establishes a quantitative link between ultrasound imaging and the disease state of the lungs. We hypothesize that ultrasound beamforming techniques that are designed specifically for the lung and its complex reverberation physics will generate higher quality images, improved clinical interpretability, and diagnostic capabilities. We will develop acoustical simulation tools and simulations of the human body and lung disease that are experimentally calibrated to accurately represent the relevant reverberation physics, such as A-line and B-line artefacts. Spatial coherence beamformers, which rely on reverberation as a source of contrast and machine learning beamformers will be designed and optimized to detect lung disease. These beamformers will be implemented on a programmable scanner and compared to conventional B-mode imaging. If successful, this proposal will yield ultrasound imaging methods that are more sensitive to lung disease, with clearer clinical interpretability, that can be deployed in current ultrasound imaging systems.