The aim of our study was to evaluate the value of a pictorial atlas of 123I FP-CIT SPECT images for aid in the visual diagnosis.
Patients, materials, methods: Sixty patients, of whom 20 were clinically diagnosed as 'non-parkinsonian' and 40 as having Parkinson's disease or any related disorder, were included in the study. An atlas consisting of 12 123I FP-CIT SPECT images was constructed first. Validity of the atlas was investigated by performing a receiver operating characteristic (ROC) analysis with the clinical diagnosis as the gold standard. The remaining 48 SPECT images were visually assessed twice by 5 observers, first with and secondly without consulting the atlas, or vice versa. The added value of the atlas was investigated by comparing the diagnostic accuracy and the interobserver variability for both methods.
Results: ROC analysis performed on the atlas yielded an area under the curve of 1 for a threshold discriminating between clinically non-parkinsonian and parkinsonian patients that was situated between image 4 and 5 of the atlas. For the diagnostic accuracy, we found that the area under the ROC curve was systematically higher if observers had access to the atlas compared to when they had not (Wilcoxon's test, p<0.05). Also, the interobserver variability was significantly lower when observers used the atlas when compared to when they did not (p = 0.05).
Conclusion: Diagnostic accuracy was significantly higher and interobserver variability significantly lower if observers had access to the atlas compared to when they had not. Hence, having a pictorial atlas available may facilitate the visual assessment of 123I FP-CIT SPECT scans.