Relating lung pathology to ultrasound imaging: An experimentally validated simulation approach

Abstract

Lung ultrasound (LUS) imaging can be highly sensitive to disease. However, lung imaging depends on reverberation that occurs at the lung interface, which is complex and upends the conventional time-space relationship in delay-and-sum beamforming resulting in images that require the interpretation of artifacts. Establishing a clear link between ultrasound images and underlying alveolar or fibrotic state of the lung could improve the diagnostic accuracy and clinical deployment of lung ultrasound and potentially establish LUS as a gold-standard imaging modality. Here, it is shown how histology-derived acoustical maps of the lung, Visible Human maps of the abdomen, and Fullwave simulations of ultrasound propagation can accurately model the multiple scattering physics at the lung interface. Lung B-mode images are generated based on the first principles of propagation and multiple scattering and they are compared to clinical imaging. In silico modifications of the aeration/porosity and the fluid-to-tissue ratio in the lung parenchyma are related to the corresponding changes in B-mode images. Additionally, the patterns of superficial/subpleural air inclusions were analyzed and mapped to corresponding B-mode image markers (white lung, single and multiple B-lines, A-lines). This establishes a validated framework for quantitative imaging of lung disease and the development of LUS-specific beamforming.