Image Processing Framework for Pleural Line (A-Line) Detection in Video Lung Ultrasonography

The implementation of Point-of-care Ultrasound (POCUS) for lung imaging has significant potential in the diagnostic of lung abnormalities through the detection of artifacts in lung ultrasonography (LUS), i.e., pleural line (A-line) and vertical comet-tail artifact (B-line). Detecting the pleural line pattern is an essential feature for further lung diagnosis based on machine learning. A healthy lung correlates with the regular repeated horizontal A-line with a fixed distance between the lines and, ideally, produces a higher intensity. This preliminary work focuses on developing an image processing framework for automatic pleural line (A-line) detection in time series B-mode ultrasound images (video) in a healthy subject as an early stage of further lung image interpretations COVID-19 pneumonia patients. The proposed scheme is based on a top-hat morphological grayscale filter with a texture structure element. An adaptive low pass filter that considers local shape parameter is used to suppress noise and keep the curve line information related to the pleural line. The proposed scheme is evaluated for an open dataset of video LUS. The proposed method can successfully detect the pleural line automatically for typical video LUS acquired using a curved transducer.