Improved visual [(123)I]FP-CIT SPECT interpretation for evaluation of parkinsonism by visual rating of parametric distribution volume ratio images.

AIM: Imaging of presynaptic dopamine transporters (DAT) by single-photon emission computed tomography (SPECT) and [(123)I]FP-CIT is an established method for differentiating between neurodegenerative and non-neurodegenerative parkinsonism. Whereas a region-of-interest (ROI) analysis is the method of choice for analyzing [(123)I]FP-CIT SPECT studies, visual image interpretations can also provide highly accurate results. The present study was undertaken to validate a visual reading system for parametric volume of distribution (DVR) [(123)I]FP-CIT SPECT images that combines the quantitative nature of ROI analyses and the simplicity of visual readings.
METHODS: A 9-step linear visual rating template for semi-quantitative DVR ratings of caudate nucleus and putamen was developed (VRDVR). The conventional 4-step visual reading system that is mainly based on the [(123)I]FP-CIT uptake pattern was used for comparison (VRP method). Six independent observers retrospectively rated the [(123)I]FP-CIT scans of 30 consecutive parkinsonism and tremor patients (N.=16 neurodegenerative, N.=14 non-neurodegenerative) using VRDVR and VRP. In addition, a highly trained investigator performed manual ROI analyses.
RESULTS: The ROI analysis provided complete separation of both patient groups by comparing the lower DAT binding of both putamina (i.e., putamen contralateral to clinically most affected side in neurodegenerative parkinsonism). Using VRP, the two most experienced observers correctly classified all patients while 20 false-positive ratings occurred in the less experienced observers (mean area under the receiver operating characteristic curve [AUCROC] of all observers 0.93±0.07). The VRDVR ratings of the two most experienced observers did not overlap between patient groups, although at different VRDVR score cut-offs. Using the same VRDVR score cut-off for all observers, only six false-negative and one false-positive ratings occurred in total (AUCROC 0.99±0.01). Inter-observer agreement was good for VRP and VRDVR. Moreover, semi-quantitative VRDVR and quantitative ROI analyses showed a strong correlation in all observers (Spearman's rho, 0.85-0.91).
CONCLUSIONS: The proposed VRDVR method offers a very promising visual analysis method for [(123)I]FP-CIT SPECT studies in parkinsonism. The accuracy of VRDVR readings was found to be superior to conventional VRP, while it provided a diagnostic accuracy in less experienced observers that is comparable to manual ROI analyses by a highly trained investigator.