Novel Quantitative Lung Ultrasound Spectroscopy Approach for Diseases Classification

Lung ultrasound (LUS) is known to be a non-ionized and real-time imaging modality capable of detecting injured areas on the lung surface. However, LUS application is qualitative and based on the subjective interpretation of imaging artifacts. To enhance reproducibility and reduce subjectivity, moving to quantitative approaches is needed. In this study, we present a quantitative approach based on the multifrequency analysis of three regions defined by the pleural-line landmark. Precisely, from a group of 17 patients affected by Pulmonary Fibrosis (PF), Pulmonary Fibrosis Exacerbated (PFE), and Chronic Obstructive Pulmonary Disease (COPD), radiofrequency (RF) data were acquired with a linear probe. From each RF frame, the pleural line was manually selected to divide the image into three regions, i.e., pre-pleura, pleura, and post-pleura. The total intensity of these regions and the ratios pre-pleura/pleura, pleura/post-pleura are analyzed, along with their frequency trend (native frequency and bandwidth). The features obtained and their significance in discriminating lung diseases are analyzed through binary decision tree classifiers. The results show how the spectral signature of different regions of the image can differentiate COPD with PF and within PF can stratify, among different levels of PF (PF and PFE).