Sensitivity, Specificity and Test-Retest Reliability Comparison of a Novel Real-Time and four Offline Resting-State fMRI Analysis Pipelines

Motivation: Standardization of resting-state fMRI (rsfMRI) data analysis is an increasingly important goal for achieving clinical acceptance. A performance comparison of widely used seed-based rsfMRI analysis pipelines has not been reported. Goal(s): We compare a real-time rsfMRI analysis pipeline (TurboFIRE) with AFNI, CONN, C-PAC and IClinfMRI. Approach: Human Connectome Project rsfMRI data (25 subjects) and high-speed rsfMRI data (15 controls and 14 brain tumor patients) were analyzed to assess 8 performance metrics of sensitivity, specificity and test-rest reliability. Results: Considerable variability of performance metrics was measured between pipelines with real-time rsfMRI analysis pipeline results being within the range of the offline analysis pipeline results. Impact: This study emphasizes the need to further standardize rsfMRI analysis pipelines. It also shows that real-time rsfMRI analysis now approaches the sensitivity, specificity and test-retest reliability of state-of-the-art offline analysis pipelines.