Second-order grey-scale texture analysis of pleural ultrasound images to differentiate acute respiratory distress syndrome and cardiogenic pulmonary edema

Abstract

Discriminating acute respiratory distress syndrome (ARDS) from acute cardiogenic pulmonary edema (CPE) may be challenging in critically ill patients. Aim of this study was to investigate if gray-level co-occurrence matrix (GLCM) analysis of lung ultrasound (LUS) images can differentiate ARDS from CPE. The study population consisted of critically ill patients admitted to intensive care unit (ICU) with acute respiratory failure and submitted to LUS and extravascular lung water monitoring, and of a healthy control group (HCG). A digital analysis of pleural line and subpleural space, based on the GLCM with second order statistical texture analysis, was tested. We prospectively evaluated 47 subjects: 16 with a clinical diagnosis of CPE, 8 of ARDS, and 23 healthy subjects. By comparing ARDS and CPE patients' subgroups with HCG, the one-way ANOVA models found a statistical significance in 9 out of 11 GLCM textural features. Post-hoc pairwise comparisons found statistical significance within each matrix feature for ARDS vs. CPE and CPE vs. HCG (P ≤ 0.001 for all). For ARDS vs. HCG a statistical significance occurred only in two matrix features (correlation: P = 0.005; homogeneity: P = 0.048). The quantitative method proposed has shown high diagnostic accuracy in differentiating normal lung from ARDS or CPE, and good diagnostic accuracy in differentiating CPE and ARDS. Gray-level co-occurrence matrix analysis of LUS images has the potential to aid pulmonary edemas differential diagnosis.

Keywords: Acute respiratory failure; Artificial intelligence; Computer aided diagnosis; Heart failure; Lung ultrasonography; Quantitative lung ultrasonography.

Fig. 1

In second-order statistical texture analysis, information on texture is based on the probability of finding a pair of grey-levels at random distances and orientations over an entire image. This is done through computing Grey-Level Co-Occurrence Matrices (GLCMs). The entries in a GLCM are the probability of finding a pixel with grey-level I, having set a distance d and angle θ from a pixel with a grey-level j, that is: P(i, j:d, θ). An essential component of this framework is pixel connectivity, where each pixel has eight nearest-neighbours connected to it, except at the periphery. As a result four GLCMs are required to describe the texture content in the horizontal (PH = 0°), vertical (PV = 90°) right (PRD = 45°) and left-diagonal (PLD = 135°) directions. The information extracted from these matrices can be used for computing textural features, specifically designed for this purpose which are sensitive to specific elements of texture. Panel a: In the image, a local zoom of a healthy pleural line area highlights that brighter (white) regions are present against a “darker” (light grey) background that results in high positive “Cluster Shade” values. Panel b: shows a local zoom in the pleural line area of an acute cardiogenic pulmonary edema subject (globally looking similar to a healthy one to the human eye) presents darker (light/dark grey) regions against a lighter background. This results in negative “Cluster Shade” values. Moreover, a local zoom of the pleural line area shows small regions with uniform dark grey intensity resulting in low “Correlation”. Panel c: in this image, local zoom of an ARDS pleural line area shows large regions with uniform dark grey intensity resulting in high “Correlation”

Fig. 2

ROC curves of texture features in differentiating acute pulmonary edema and acute respiratory distress syndrome ultrasound patterns