Multidetector computed tomography (MDCT) in the diagnosis of pulmonary embolism: interobserver agreement among radiologists with varied levels of experience

Abstract

Purpose: To assess the interobserver variability of radiologists with varied levels of experience in the interpretation of multidetector computed tomography (MDCT) pulmonary angiographies. Material and Methods: Review of CT pulmonary angiographies performed on patients included in a diagnostic study evaluating a decision-based algorithm for diagnosing pulmonary embolism (PE). Five radiologists, three board-certified general radiologists and two radiology trainees with 2 years' experience, participated in the study. Results: According to the consensus reading, PE was present in 91 (31%) and absent in 194 (67%) patients, while in five patients (1.7%) the interpretations were regarded as equivocal. The per-patient agreement on the diagnosis of PE achieved by each of the four readers compared to the consensus reading was very good (κ range 0.85–0.92), but peripheral emboli were missed in four to six patients by three of four observers. The agreement on the most proximal level of PE (per-proximal level) assessed by mean κ value was 0.83 (κ range 0.68–0.91) for the detection of proximal emboli, 0.61 for segmental emboli (κ range 0.40–0.80), and 0.38 for emboli in the subsegmental vessels (κ range 0.0–0.89). Conclusion: The overall agreement on the diagnosis of PE by MDCT for general radiologists and radiology trainees is very good, and we therefore believe that the initial management of patients with suspected PE could be based on the preliminary assessment performed by on-call radiologists with 2 years of experience.