Dynamic time warping outperforms Pearson correlation in detecting atypical functional connectivity in autism spectrum...

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Clinical Psychology

Background:

• Resting-state functional magnetic resonance imaging (rsfMRI) assesses brain function by measuring correlated blood-oxygen-level-dependent (BOLD) signal fluctuations.
• Functional connectivity (FC) is typically quantified using Pearson correlation, assuming linear relationships between BOLD time series.
• Linear methods may underestimate true FC due to non-linear dynamics, hemodynamic differences, or signal lags, particularly in clinical populations.

Purpose of the Study:

• To investigate the efficacy of non-linear dynamic time warping (DTW) as an alternative to Pearson correlation for quantifying rsfMRI functional connectivity (FC).
• To compare the robustness, reliability, and sensitivity of DTW versus Pearson correlation in diverse age groups and clinical populations, including autism spectrum disorders (ASDs).
• To explore the relationship between DTW-derived FC, clinical measures, and physiological noise.

Main Methods:

• Analysis of rsfMRI data from three cohorts: children, adolescents, and adults, with and without autism spectrum disorders (ASDs).
• Quantification of functional connectivity (FC) using both traditional Pearson correlation and non-linear dynamic time warping (DTW).
• Assessment of robustness to global signal regression (GSR), test-retest reliability, sensitivity to task-related FC changes, and association with clinical and physiological measures.

Main Results:

• Dynamic time warping (DTW) demonstrated greater robustness to global signal regression (GSR) and higher test-retest reliability compared to Pearson correlation.
• DTW was more sensitive to task-related changes in functional connectivity (FC) and detected more group differences in individuals with autism spectrum disorders (ASDs).
• DTW-derived FC estimates showed stronger associations with ASD symptom severity and executive function, whereas Pearson correlation was more linked to respiration.

Conclusions:

• Non-linear methods, specifically dynamic time warping (DTW), offer a more accurate and reliable approach to estimating resting-state functional connectivity (FC) from rsfMRI data.
• DTW is particularly advantageous for studying clinical populations, such as those with autism spectrum disorders (ASDs), where altered hemodynamics or neurovascular coupling may exist.
• The findings suggest that DTW enhances the sensitivity and validity of rsfMRI analyses, providing deeper insights into brain function and its relationship to clinical conditions.