Interobserver reproducibility in the interpretation of 99mTc-labelled white blood cell scintigraphic images

This study was performed to investigate the interobserver reproducibility of the interpretation of Tc-labelled white blood cell scans combined or not with other nuclear medicine procedures.

Twenty nuclear medicine physicians working in Belgium received clinical data and scintigraphic images from 10 patients suspected of suffering from various infectious diseases. They had to choose, for each patient, one answer among 'high probability', 'intermediate probability' and 'low probability' of infection. In a first step the level of agreement, defined as the highest percentage obtained among the three proposed answers, was calculated for each case; complete agreement was arbitrarily defined when 80% or more of the observers gave the same answer. In a second step, a numerical score was given for each answer and for each observer. The scores were defined as 0 for the 'low probability' answer, 1 for the 'intermediate probability' and 2 for the 'high probability' answer. The scores of each observer were summed up allowing obtaining of a cumulated score for each observer; observers were thereafter classified as a function of their own cumulated score.

An agreement of 80% or more was observed only in three patients. The cumulated scores of the observers were between 4 and 16, demonstrating that some observers were more sensitive than others.

Interobserver reproducibility in the interpretation of the white blood cell scan was poor; several factors could explain these results, the most frequent being both the poor specificity of the labelled white blood cell scan and different concepts in interpreting combined nuclear medicine procedures in several situations. As observers received the anonymized overall results, this study may have a favourable impact on continuing education in medical imaging.