V15. Ex post facto structural connectome analysis in ALS at multicenter level: Analysis of over 400 data sets from 8 centers

Abstract

Introduction Diffusion tensor imaging (DTI)-based metrics are increasingly used for analyzing ALS-associated whitematter alteration patterns and were included inthe Neuroimaging Society in Amyotrophic Lateral Sclerosis (NiSALS) concept ( Turner et al., 2011 ). The objective of this multicenter study was to assess structural connectivity in ALS at a large sample size to address the challenges of DTI data analysis from multiple study sites. Methods Four-hundred and ninety DTI data sets from patients with ALS ( N = 268) and controls ( N = 222) were collected from 8 study centers (Dublin, Ireland; Edinborough, UK; Jena, Germany; Miami, US; Milan, Italy; Oxford, UK; Rostock, Germany, Ulm, Germany). Data were obtained by different MRI-systems and by use of different DTI-protocols. In a first step, comparability of data with the aim of pooling was tested by a statistical analysis of fractional anisotropy (FA) in controls’ data. All analyses were performed by use of the Tensor Imaging and Fiber Tracking (TIFT) software. Statistical comparisons in terms of average FA-values ( Muller et al., 2013 ) were performed for the controls’ groups of the different centers. In a consecutive step, the calculated 3D correction matrices were then applied to the corresponding ALS patient subgroups of the different centers. Results Data collection and data quality control resulted in 442 DTI data sets (253 ALS patients and 189 controls) useful for this study, i.e. 48 data sets had to be excluded. All data samples of all centers showed a characteristic pattern (FA decrease along the corticospinal tracts) for comparison at the group level. Inter-center pooling of data showed ALS affectations in the CST (“horseshoe” configuration) as well as affectations in regions that have recently been reported to be relevant in ALS (frontocortical whitematter regions, hippocampal regions, as well as midbrain and brainstem) ( Kassubek et al., 2014 ). Conclusion This large scale multicenter NiSALS study investigated solutions to challenges in the process of pooling MRI data recorded at various study centers in ALS. This approach is of utmost importance in order to establish MRI-based techniques as read-outs both fornatural history assessment and for potential upcoming disease-modifying multicenter studies in ALS ( Roskopf et al., in press ).