INVESTIGATING RELIABILITY AS A TOOL FOR PATIENT-SPECIFIC PIPELINE SELECTION

Abstract

Introduction: Functional neuroimaging plays a key role in pre-surgical planning and improving surgical outcomes. My team has developed ROCr (Receiver Operating Characteristic reliability) - a tool to assess intra-session reliability of functional neuroimaging data. Objective: Investigate ROCr as a tool for automated, patient-specific analysis. Methods: Ten subjects underwent three MEG (magnetoencephalography) imaging sessions of right MNS (median nerve stimulation). Twelve pipelines were used to produce activation maps for each dataset, and ROCr was used to assess the reliability of these maps. Various tests were done on the results to assess ROCr. Results and Discussion: Within subjects, the pipeline with the highest reliability varied across days, likely because the inter-session data variance was larger than assumed. Using the pipeline with highest reliability resulted in lower inter-session variability. Further analysis of the outcome measures gave new insight into how to improve use of ROCr in future studies and clinical applications.