A Novel Empirical Wavelet Transform Approach for Classification of Radiofrequency Lung Ultrasound Signals Applied to Diagnosis of Lung Diseases

LUS analysis mainly focuses on evaluating imaging artifacts, especially the vertical ones. This analysis remains subjective, hence poorly reproducible. To mitigate these flaws, quantitative approaches based on analyzing the vertical artifacts' spectral signatures have been presented. Although previous studies proved the capabilities of LUS spectroscopy in discriminating lung diseases, the diagnostic insights carried by vertical artifacts need to be investigated more deeply. In this study, we present a novel signal processing technique applied to vertical artifact patterns to differentiate lung diseases. Specifically, we work with multifrequency radiofrequeny (RF) data acquired from a group of 114 patients, however, we selected 31 patients acquired with a linear probe and affected by cardiogenic pulmonary edema (CPE) and pneumonia. From these data, we extract two sets of statistical features using two different approaches.