Establishing Consensus in Categorizing Resting State fMRI Brain Networks: An Inter-Rater Reliability Study (2024)

Undergraduates: Sriya Darsi, Maithili Kulkarni

Faculty Advisor: Varina Boerwinkle
Department: Neurology

---

Successful clinical use of resting state functional magnetic resonance imaging (rs-fMRI) analysis hinges upon accurate categorization of intrinsic brain networks, yet variability in interpretation among clinicians poses a challenge. We present an inter-rater reliability study aimed at establishing a gold standard for categorizing rs-fMRI-derived brain networks. Leveraging an independent component analysis (ICA) pipeline, raters categorized networks using an internal report generation tool, reflecting interpretations across nine anatomical regions and four specific categories including “normal”, “atypical”, “atypical resting state”, and “noise”. Preliminary findings reveal variability among raters, underscoring the need for a consensus framework to ensure consistency in clinical interpretation. Our analysis revealed moderate to substantial concordance among raters regarding the classification of “atypical” and “noise,” whereas agreement was minimal to absent for the “other” category. In most instances, a sole rater assigned the “other” classification for a participant's data, while the remaining raters consistently concurred on decisions related to “atypical” or “noise,”. Our study highlights the importance of interdisciplinary collaboration in refining categorization protocols to enhance clinical reliability and reproducibility in neurological research.