Large-scale augmented Granger causality (lsAGC) for enhanced analysis of brain connectivity in autism spectrum disorder

We introduce large-scale augmented Granger causality (lsAGC) as a novel method for analyzing directed connectivity in high-dimensional neural systems. The lsAGC algorithm integrates invertible dimension reduction with time-series augmentation to overcome the challenges faced by current methods in analyzing connectivity within complex systems. This multivariate approach effectively determines the influence of each time series on others within the context of all time-series in a dynamic system. We applied lsAGC to resting-state fMRI data from the Autism Brain Imaging Data Exchange II (ABIDE II) dataset, with a focus on data from the Olin Neuropsychiatry Research Center. Our method shows improved accuracy in differentiating individuals with autism spectrum disorder (ASD) from healthy controls (HLT). Quantitative evaluation shows that lsAGC achieves a higher AUC of 0.940±0.074 compared to competing methods: correlation (0.890±0.100, −5.32%), local models (0.905 ± 0.108, −3.72%), and antisymmetric correlation (0.882 ± 0.102, −6.17%). Statistical analysis using Wilcoxon signed rank tests confirms significant improvements of lsAGC over all competing methods (p ⪅ 10⁻⁴ for all comparisons). Furthermore, lsAGC shows lower entropy values (ASD: 1.563 ± 0.150, HLT: 1.543 ± 0.216) compared to other methods, suggesting that it captures more discriminative features. These results indicate the potential of lsAGC to improve our understanding of brain connectivity patterns in ASD and other neurological conditions. By extracting sparse connectivity matrices, lsAGC can be particularly useful for network analysis in complex systems and could be applicable to clinical fMRI analysis in future research, such as targeting disease-related classification or regression tasks in clinical data.