Intra- and interobserver variability of board-certified veterinary radiologists and veterinary general practitioners for pulmonary nodule detection in standard and inverted display mode images of digital thoracic radiographs of dogs

Abstract

To determine intra- and interobserver variability of 2 veterinary radiologists and 2 veterinary general practitioners for detection of pulmonary nodules in standard and inverted (reversed grayscale) displays of digital thoracic radiographs of dogs. Evaluation study. 114 sets of 3-view (right lateral, left lateral, and ventrodorsal or dorsoventral views) digital thoracic radiographs from 114 dogs. 2 experienced board-certified veterinary radiologists and 2 experienced veterinary general practitioners individually evaluated 114 randomized sets of radiographs. Pulmonary nodules were present in radiographs of 60 of 114 dogs. Each reviewer examined all images in standard or inverted display mode and scored nodule detection on a confidence scale of 1 to 5. After ≥ 2 months, the same individuals evaluated the same images in the remaining display mode. Intraobserver agreement for each display mode was determined via a κ statistic; results between the 2 groups of reviewers were compared via receiver operator curve analysis. There was no significant intraobserver variability in pulmonary nodule detection between the 2 display modes. Detection accuracy for board-certified radiologists was significantly greater than that of veterinary general practitioners for both display modes. Near-perfect intraobserver agreement was detected between the 2 display modes for board-certified radiologists, whereas moderate to slight intraobserver agreement was detected for the veterinary general practitioners. Detection of pulmonary nodules in digital thoracic radiographs was comparable, whether a standard or inverted mode was used for evaluations. However, the board-certified radiologists had greater detection accuracy than did veterinary general practitioners.