Lung Ultrasound for Pleural Line Abnormalities, Confluent B-Lines, and Consolidation: Expert Reproducibility and a Method of Standardization.

Abstract

Discrete B-lines have clear definitions, but confluent B-lines, consolidations, and pleural line abnormalities are less well defined. We proposed definitions for these and determined their reproducibility using COVID-19 patient images obtained with phased array probes. Two raters collaborated to refine definitions, analyzing disagreements on 107 derivation scans from 10 patients. Refined definitions were used by those raters and an independent rater on 1260 validation scans from 105 patients. Reliability was evaluated using intraclass correlation coefficients (ICC) or Cohen's kappa. The agreement was excellent between collaborating raters for B-line abnormalities, ICC=0.97 (95% confidence interval [CI] 0.97-0.98) and pleural line to consolidation abnormalities, ICC=0.90 (95% CI 0.87-0.92). The independent rater's agreement for B-line abnormalities was excellent, ICC=0.97 (95% CI 0.96-0.97) and for pleural line to consolidation was good, ICC=0.88 (95% CI 0.84-0.91). Agreement just on pleural line abnormalities was weak (collaborators, κ=0.54, 95% CI 0.48-0.60; independent, κ=0.54, 95% CI 0.49-0.59). With proposed definitions or via collaboration, overall agreement on confluent B-lines and pleural line to consolidation abnormalities was robust. Pleural line abnormality agreement itself was persistently weak and caution should be used interpreting pleural line abnormalities with only a phased array probe.