Interobserver variability, and visual and quantitative parameters of (123)I-FP-CIT SPECT (DaTSCAN) studies

Abstract

Objective: To evaluate the degree of interobserver agreement in the visual interpretation of (123)I-FP-CIT studies and to investigate for potential associations between visual and semi-quantitative parameters.

Methods: Eighty-nine (123)I-FP-CIT studies were blindly reviewed by 3 independent observers: a consultant, a resident doctor and a radiographer. They classified every study as either "normal" or "abnormal" and assigned visual (123)I-FP-CIT uptake scores (2: normal, 1: reduced and 0: no uptake) in basal ganglia nuclei (right and left putamina and caudate nuclei) on every scan. Striatal (123)I-FP-CIT binding ratios were calculated using crescent-ROI software. The interobserver agreement for the interpretation of studies and for visual score assignment was evaluated by means of κ statistics. We investigated for associations of binding ratios with visual scores and clinical parameters; patients' clinical diagnoses served as the reference standard.

Results: There was excellent interobserver agreement (κ 0.89-0.93) in classifying studies as "normal" or "abnormal" and fine agreement in assignment of visual scores (κ 0.71-0.80 for putamina and 0.50-0.79 for caudate nuclei). Nuclei with scores of 1 and 0 showed significantly reduced binding ratios (about 30 and 50%, respectively) compared with the nuclei scored as 2. ROC analysis indicated the optimal cutoff point of striatal binding ratio at 3.8 (sensitivity 98.5%, specificity 95%) for the detection of parkinsonian syndromes. Striatal binding ratios were negatively associated with age in normal subjects and disease duration in Parkinson's disease patients.

Conclusion: Visual interpretation of (123)I-FP-CIT studies showed very good interobserver agreement. We found significant associations among visual, semi-quantitative and clinical parameters.