Visual versus automated analysis of [I-123]FP-CIT SPECT scans in parkinsonism

Abstract

The clinical evaluation of dopamine transporter (DAT) SPECT scans typically relies on visual analysis in combination with an automated semi-quantitative method. The interpretation of the results may be difficult in cases that show disagreement between the two methods on the borderline of abnormality. The frequency and clinical characteristics of such cases are unclear. Automated semi-quantitative analyses and independent visual analyses by two experienced nuclear medicine physicians and four inexperienced raters were performed for 120 patients with clinically uncertain parkinsonism scanned with brain [I-123]FP-CIT SPECT. Agreement was evaluated with kappa statistics. The clinical characteristics of patients who had discrepant findings between the two analysis methods were investigated. The expert raters outperformed nonexperts in terms of agreement between visual and automated analyses (kappa = 0.66, 0.72 vs. 0.23-0.54) and between raters (kappa = 0.81 vs. 0.44-0.63). Twelve patients showed discrepant findings between the visual and automated analyses. These patients were older compared to other patients (p = 0.023), had 17.6 % lower mean striatal tracer binding compared to normal scans (p = 0.003) and 62.7 % higher compared to abnormal scans (p < 0.001). After a minimum of 4.5 years of clinical follow-up, none of these patients developed neurodegenerative parkinsonism. Clinical DAT SPECT scans show discrepancies between visual and automated analyses in 10 % of cases. The patients with discrepant findings are older, show normal to slightly abnormal tracer binding, and importantly, do not develop neurodegenerative parkinsonism syndromes. Visual analyses by experienced raters are reliable, but the diagnostic accuracy in discrepant cases can be improved by an automated method.