The underestimation of segmental defect size in radionuclide lung scanning

Criteria used to place ventilation-perfusion lung scans into categories with different probabilities for pulmonary embolism depend largely on the size and anatomical distribution of defects recognized. These criteria assume that actual segmental defects appear segmental on the lung scan. This study examined the accuracy with which four experienced observers were able to estimate the size of defects of known anatomical location and size, using images of segmental defects in ventilation produced with a bronchoscopic technique and 81mKr. Of the 24 segmental defects produced in this study, 17% were interpreted as being < 25% of a segment; 23% were interpreted as being 25%-50% of a segment; 17% were interpreted as 50%-75% of a segment; 40% were interpreted as being 75%-100% of a segment and 4% were interpreted as being > 100% of a segment. Intra- and interobserver agreement as assessed by the Kappa statistic varied with the number of size categories used but was generally poor. Underestimation of defect size observed in this study may explain why many patients with pulmonary embolism do not have high probability scans. We conclude that the subjective impression of the size of a defect on a lung scan is an unreliable indication of a defect's true segmental or subsegmental nature and that scoring systems based on these criteria should be viewed with caution.