Our pipeline consists of a hand-crafted preprocessor and a neural network classifier. We applied transformations on the physiologic signals to gain features in both time-and frequency domains. The proposed algorithm was trained on 994 annotated records of polysomnographic signals. Most of the features were generated from the EEG signal such as power spectral density, and entropy. We extracted features from the EOG, EMG, airflow, and ECG signals too. All the features were normalized. These 68 features were resampled in 21 non-continuous moments around the current timestamp, and fed into a 3layer neural network in order to assign a probability of arousal at each second. Arousal samples were enriched during training to battle data imbalance. Additional (auxiliary) losses can guide the network to learn high-level concepts, even though they will not be evaluated. We used sleep stages as additional training targets, which were easier to learn than arousals despite being multi-class. This approach slightly increased arousal AUPRC. Our submitted results for the entire test set were evaluated: AUPRC=0.42. Our 10-fold cross validation results for the AUPRC are the following: [

Hierarchical counterstream via feedforward and feedback interactions is a major organizing principle of the cerebral cortex. The counterstream, as a topological feature of the network of cortical areas, is captured by the convergence and divergence of paths through directed links. So defined, the convergence degree (CD) reveals the reciprocal nature of forward and backward connections, and also hierarchically relevant integrative properties of areas through their inward and outward connections. We asked if topology shapes large-scale cortical functioning by studying the role of CD in network resilience and Granger causal coupling in a model of hierarchical network dynamics. Our results indicate that topological synchronizability is highly vulnerable to attacking edges based on CD, while global network efficiency depends mostly on edge betweenness, a measure of the connectedness of a link. Furthermore, similar to anatomical hierarchy determined by the laminar distribution of connections, CD highly correlated with causal coupling in feedforward gamma, and feedback alpha-beta band synchronizations in a well-studied subnetwork, including low-level visual cortical areas. In contrast, causal coupling did not correlate with edge betweenness. Considering the entire network, the CD-based hierarchy correlated well with both the anatomical and functional hierarchy for low-level areas that are far apart in the hierarchy. Conversely, in a large part of the anatomical network where hierarchical distances are small between the areas, the correlations were not significant. These findings suggest that CD-based and functional hierarchies are interrelated in low-level processing in the visual cortex. Our results are consistent with the idea that the interplay of multiple hierarchical features forms the basis of flexible functional cortical interactions.

No abstract