A Critical Assessment of Directed Connectivity Estimates with Artificially Imposed Causality in the Supramammillary-Septo-Hippocampal Circuit

Algorithms for estimating directed connectivity have become indispensable to further understand the neurodynamics between functionally coupled brain areas. The evaluation of directed connectivity on the propagation of brain activity has largely been based on simulated data or toy models, where various hidden properties of neurophysiological data may not be fully recapitulated. In this study, directionality was unequivocally manipulated in the freely moving rat in a unique dataset, where normal oscillatory interactions between the supramammillary nucleus (SuM) and hippocampus (HPC) were attenuated by temporary medial septal (MS) inactivation, and replaced by electrical stimulation of the fornix to evaluate the performance of several directed connectivity assessment methods. The directed transfer function, partial directed coherence, directed coherence, pair-wise Geweke-Granger causality, phase slope index, and phase transfer entropy, all found SuM to HPC theta propagation when the MS is inactivated, and HPC activity was driven by peaks of simultaneously recorded SuM theta. As expected from theoretical expectations and simulated data, signal features including coupling strength, signal-to-noise ratio, and stationarity all weakly affected directed connectivity measures. We conclude that all the examined directed connectivity estimates correctly identify artificially imposed uni-directionality of brain oscillations in freely moving animals. Non-auto-regressive modeling based methods appear to be the most robust, and are least affected by inherent features in data such as signal-to-noise ratio and stationarity.