63 consecutive patients with a clinical suspicion of PE who underwent V/P-SPECT were included in this retrospective study. Images were randomly ordered both for standard as well as for software-assisted reading using V/P-quotients. Studies were read independently by 2 experienced and 2 inexperienced raters. Diagnostic performance and observer agreement of all readers and both reading methods were determined.
Expert observers consistently achieved a high diagnostic accuracy both in conventional as well as in software-assisted reporting (sensitivity: 0.94 vs. 0.94, specificity: 0.96 vs. 0.97, LR: 17.32 vs. 28.86, LR stayed constant at 0.06). For inexperienced readers, diagnostic performance improved: sensitivity raised from 0.74 to 0.85 and specificity from 0.86 to 0.95, LR raised from 5.20 to 15.69, LR decreased from 0.31 to 0.16. Inter-rater reliability (Fleiss’ κ) improved from 0.63 to 0.86 by using V/P quotient.
Benefit from a software-tool that calculates V/P-ratio automatically is only small when used by experienced physicians If inexperienced readers use the software, the diagnostic accuracy increases. Images generated by automated calculation of V/P-mismatches are easy to read and their use might help to standardize and objectify interpretation of V/P-SPECT in the diagnosis of PE.
Thieme. All rights reserved.