Measurement of inter- and intra-observer variability in the routine clinical interpretation of brain 18-FDG PET-CT.

ABSTRACT

OBJECTIVE: To objectify and quantify inter- and intra-observer variability of brain 18-FDG PET-CT interpretation in the context of cognitive and functional impairment amongst the elderly. METHODS: 25 patients underwent brain 18-FDG PET-CT for investigation of dementia/MCI and frail elderly patients. Three observers interpreted studies in two forms standardised datasets reconstructed by an outside observer and individualised reconstructions. Observers graded regional 18-FDG uptake in 11 brain areas and gave overall impressions on studies as pathological/normal. One observer repeated this process following a 3-month interval. The Kappa statistic was used to calculate inter- and intra-observer agreement on grading of regional 18-FDG uptake and overall impressions of studies as pathological/normal. RESULTS: Moderate inter-observer agreement was observed across standardised and individualised dataset reconstructions when 11 regional brain areas were compared cumulatively and overall impressions on studies were given as pathological vs normal. Higher levels of inter-observer agreement were found when comparing high versus low grading of regional uptake and when reporting standardised reconstructions. Intra-observer agreement between standardised vs individualised dataset reconstructions were moderate-to-fair across 11 brain regions cumulatively. Temporal intra-observer agreement of individualised dataset reconstructions comparing normal vs pathological opinions showed strong agreement (κ = 0.884 [95 % CI 0.662; 1.000)]. CONCLUSION: Despite a strong agreement in final diagnosis, this study demonstrates a moderate inter- and substantial intra-observer reproducibility in reporting brain 18-FDG PET-CT. Such results suggest that the visual analysis approach is different between nuclear physicians but leads to the same final diagnosis.