Test-retest reliability of template-defined vs. manual tracing processing streams in neuromelanin-sensitive magnetic resonance imaging analysis

Abstract

Neuromelanin (NM) magnetic resonance imaging (MRI) is a relatively new, noninvasive method used as a proxy measure of midbrain dopamine function. Previous studies in schizophrenia have found evidence of enhanced signal in patients that are associated with positive psychotic symptoms. However, there are limited data available comparing methods of computing signal in substantia nigra (SN). We sought to examine the reliability and validity of manual tracing vs template-defined methods to help guide the field in identifying optimal approaches in NM imaging. NM-MRI was performed on 22 participants (18 with early psychosis (EP) and 4 healthy controls (HCs)) scanned twice over a 1–14 week period. Mean SN NM signal was calculated using template-defined and manual tracing methods to define SN ROIs. Intraclass correlation coefficients (ICCs) based on absolute agreement were calculated between test and retest. Correlations between NM signal for each method and symptoms in EP were also examined. ICCs for the template-defined were in the excellent range (.81–.85) and manual tracing methods were in the poor to fair range (−.14 to .56). The template-defined method showed a trend positive relationship with reality distortion symptoms (r = .45, p = .06) and the manual tracing method showed a significant positive relationship (r = .47, p = .05). Supplemental analyses highlighted the importance of thresholding and using the mode to compute CNR for identifying these relationships to symptomatology. While both methods showed clinical validity, the excellent reliability of the template-defined method suggests this technique is the preferred strategy for NM-MRI analysis.