Iterative Deconvolution Approach for Automatic Segmentation of Lung Ultrasound Vertical Artifacts

Abstract

Lung ultrasound (LUS) is an important imaging tool to evaluate the state of the lung surface. However, the presence of air does not allow the anatomical investigation of lungs. Indeed, clinicians currently base their analysis on the visual interpretation of imaging artifacts, such as the vertical ones, which are visualized in the image as hyper-echoic vertical artifacts and correlate with several pathologies. In this work, we present a technique aiming at automatically segmenting vertical artifacts by exploiting signal deconvolution. Specifically, we exploited the dependency of vertical artifacts on frequency, and used an iterative deconvolution technique to segment the artifacts in lung-mimicking phantoms and clinical data.