P242 Dynamic time warping distance based connectivity: A new method for resting-state FMRI functional connectivity analysis

摘要

Objectives Traditionally resting-state networks are analysed with methods implying zero-lag linear dependence between brain regions, i.e. functional connectivity strength between voxel pairs is characterized by the correlation-coefficient of the two measured signal. It is known that the shape and timing of hemodynamic response function differs between brain regions and this introduces artefacts in linear measures. Methods We proposed Dynamic Time Warping (DTW) distance to be used as an alternative similarity-measure between BOLD signals. Our method was validated in a longitudinal single-subject study where seed-based and whole-brain functional connectivity was calculated based on both DTW similarity and correlation. DTW connectivity’s sensitivity was assessed in a classification task, where a preprocessed public dataset was used: 126 subjects’ resting-state data and phenotypic information, including diagnosis for ADHD. Results The results of the single-subject measurements revealed that DTW similarity-based connectivity map calculation is more stable in multiple measurements than a correlation-based paradigm, as DTW-based connectivity strengths are similarly stable within and between sessions, while correlation yields larger variations between sessions. Furthermore, the stability of the DTW-based connectivity patterns result in significantly higher classification performance than the same classifiers trained on correlation-based features of connectivity. Discussion As DTW handles non-stationery processes, it results in stable connectivity patterns in multiple measurements, while its sensitivity for group differences is higher than correlation’s as the classification study shows. Conclusion The results demonstrate that DTW similarity is indeed an applicable and advantageous tool of resting-state functional connectivity analysis. Significance DTW-based connectivity can be efficiently used in longitudinal studies and in connectome-based classification tasks. ]]>