Diffuse to fuse EEG spectra – Intrinsic geometry of sleep dynamics for classification

Abstract

We propose a novel algorithm for sleep dynamics visualization and automatic annotation by applying diffusion geometry based sensor fusion algorithm to fuse spectral information from two electroencephalograms (EEG). The diffusion geometry approach helps organize the nonlinear dynamical structure hidden in the EEG signal. The visualization is achieved by the nonlinear dimension reduction capability of the chosen diffusion geometry algorithms. For the automatic annotation purpose, the support vector machine is trained to predict the sleep stage. The prediction performance is validated on a publicly available benchmark database, Physionet Sleep-EDF [extended] SC * (SC = Sleep Cassette) and ST * (ST = Sleep Telemetry), with the leave-one-subjectout cross validation. When we have a single EEG channel (Fpz-Cz), the overall accuracy, macro F1 and Cohen's kappa achieve 82.72%,75.91% and 76.1% respectively in Sleep-EDF SC * and 78.63%, 73.58% and 69.48% in Sleep-EDF ST * . This performance is compatible with the state-of-the-art results. When we have two EEG channels (Fpz-Cz and Pz-Oz), the overall accuracy, macro F1 and Cohen's kappa achieve 84.44%,78.25% and 78.36% respectively in Sleep-EDF SC * and 79.05%, 74.73% and 70.31% in Sleep-EDF ST * . The results suggest the potential of the proposed algorithm in practical applications.

Save Muscle Information–Unfiltered EEG Signal Helps Distinguish Sleep Stages

“…To show the usefulness of high frequency spectral information of the EEG, we consider a recently developed automatic sleep stage annotation algorithm from our group [22]. The algorithm is composed of three steps: preprocessing, unsupervised feature extraction, and learning.…”

“…Finally, we apply the well-established kernel support vector machine (SVM) [29] to learn the experts' knowledge by classifying the CCA fused features (indicated by Part 4 in Figure 1). We mention that the high frequency features associated with STFT was not considered in our previous work [22], otherwise the algorithm is the same. We refer readers with interest to that paper [22] for technical details.…”

“…Without the EMG, it might be challenging for a sleep expert, or a well-trained artificial intelligent (AI) system, to distinguish REM and N1. This fact has been well reflected in several state-of-the-art automatic sleep stage classification algorithms based on the single channel EEG signal [18][19][20][21][22]. As we can identify from those papers, the classification accuracy of N1 and REM is usually limited.…”

Alternating Diffusion Map Based Fusion of Multimodal Brain Connectivity Networks for IQ Prediction

“…So far the ADM has proven to be a powerful tool in voice detection from audio-visual signals [28], [29], Alzheimer's disease classification from multiple electroencephalography (EEG) signals [30], and sleep stage classification from EEG and respiration signals [31]. Here we show that the ADM can also be adapted to multimodal fMRI data.…”

Connecting dots: from local covariance to empirical intrinsic geometry and locally linear embedding

“…From the statistical viewpoint, it is a nonlinear latent space model, and the local covariance structure leads to a generalization of the Mahalanobis distance. While it has been successfully applied to different problems [29,16,31,21,15], to the best of our knowledge, except an argument on the Euclidean space setup [22], a systematic evaluation of how the algorithm works under the manifold setup, and its sensitivity to the parameter choice, is missing. Due to its importance, the first contribution of this paper is providing a quantification of EIG under the manifold setup, and discuss how the chosen parameter influences the final result.…”