MULTICENTRIC VALIDATION OF A NON-LINEAR MEASURE TO ASSESS FMRI RESTING-STATE CONNECTIVITY IN FRONTOTEMPORAL DEMENTIA

MOGUILNER SEBASTIAN; GARCIA ADOLFO; MIKULAN EZEQUIEL; HESSE EUGENIA; GARCIA CORDERO, INDIRA; MELLONI MARGHERITA; CERVETTO SABRINA; HERRERA, EDUAR; PABLOS REYES; DIANA MATALLANA; MANES FACUNDO; IBAÑEZ AGUSTIN; SEDEÑO LUCAS

Conferencia; 2018 International Conference on Frontotemporal Dementias; 2018

Although functional connectivity (FC) measures have been proposed as potential biomarkers for neurodegenerative diseases, extant research has yielded inconsistent results. However, most available studies are blind to non-linear dynamics, since they calculate FC exclusively via linear measures. To overcome this limitation, we developed a ?weighted Symbolic Dependence Dynamics? (wSDD) measure. Based on symbolic transforms, we factor in local and global temporal features of the BOLD signal to weigh a robust copula-based dependence measure by symbolic similarity, capturing both linear and non-linear associations. To test our measure, we compared it with a widely used linear connectivity metric (Pearson?s R) in its performance for discriminating patients with behavioral variant frontotemporal dementia (bvFTD) and controls, based on resting-state data. Importantly, by recruiting participants from two international centers with different MRIs recordings, we also assessed which metric proved more consistent across heterogeneous acquisition conditions. We found that, relative to Pearson, wSDD yields better and more consistent identification of canonical resting-state networks, like the salience network (a specific target of bvFTD) and the default-mode network (as a complementary control) in controls and patients. Second, machine learning analysis (via support vector machines and nearest-neighbor algorithms) revealed that wSDD yielded higher classification accuracy in the salience network than linear coupling measures to discriminate between subjects in each group. Crucially, these results were consistent in both centers, highlighting their robustness despite heterogeneous acquisition parameters and sociocultural contexts. Our findings highlight the potential of wSDD to assess fMRI-derived FC data, and, more particularly, to identify sensitive biomarkers in bvFTD and other neurodegenerative diseases. Partially supported by grants from CONICET, CONICYT/FONDECYT Regular (1170010), FONDAP 15150012, INECO Foundation, and the Inter-American Development Bank.