Cluster Confidence Index: A Streamline‐Wise Pathway Reproducibility Metric for Diffusion‐Weighted MRI Tractography. Issue 1 (21st September 2017)

• Record Type: Journal Article
• Authors: Jordan, Kesshi M.; Amirbekian, Bagrat; Keshavan, Anisha; Henry, Roland G.
• Is Part Of: Journal of neuroimaging. Volume 28:Issue 1(2018)
• Page Start: 64
• Page End: 69
• Publication Date: 2017-09-21
• Subjects: Reproducibility of results -- probabilistic tractography -- diffusion‐weighted magnetic resonance imaging -- white matter fascicle model -- open‐source software; Diagnostic imaging -- Periodicals; Nervous system -- Diseases -- Diagnosis -- Periodicals; Imagerie pour le diagnostic -- Périodiques; Système nerveux -- Maladies -- Diagnostic -- Périodiques; Imagerie médicale; Neuroimagerie; Neurologie; Système nerveux; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 616.804754
• DOI: 10.1111/jon.12467
• Languages: English
• ISSNs: 1051-2284

Abstract

ABSTRACT: BACKGROUND: Diffusion‐weighted magnetic resonance imaging tractography can be used to create models of white matter fascicles. Anatomical and pathological variability between subjects can drastically alter the tractography output, so standardizing results across a cohort is nontrivial. Furthermore, tractography methods have inherently low reproducibility due to stochasticity (for probabilistic methods) and subjective decisions, since the final fascicle model often requires a manual intervention step performed by an expert human operator to control both outliers and systematic false‐positive pathways, as defined by prior knowledge of anatomy. METHODS: We present an approach that computationally assigns a cluster confidence index (CCI) reflecting the reproducibility of that pathway in the context of a streamline dataset. This metric is a tractography algorithm‐agnostic tool that can be applied to any dataset of streamlines. RESULTS: Applications of this metric include systematic elimination of outlier streamlines using a CCI threshold and interactive filtering by CCI to facilitate manual segmentation of fascicle models. CONCLUSIONS: This method is intended to replace the application of a streamline density threshold so that outliers are eliminated based on low pathway density instead of voxel‐wise density.