Arousal labeling is one of the important methods in the diagnosis and treatment of sleep-related diseases, and are usually analyzed manually by doctors based on polysomnography (PSG) signals. In order to solve the problem of time-consuming and labor-intensive manual arousal analysis in sleep physiological signals, we propose an automatic arousal detection method using multi-model deep neural networks. Combining methods such as one-to-many formulation, LSTM, and network structure improvements, the performance of deep neural network models on clinical data set has been significantly improved, and multiple indicators have been improved (precision 86.7%, recall 86.0% and F1 86.3%). At the same time, the model parameters have been greatly streamlined, making them more efficient, laying a foundation for the application of automatic arousal detection methods on wearable sleep monitoring device signal analysis.
Paper
Full text
Automatic Arousal Detection Using Multi-model Deep Neural Network
Semantic Scholar · Computer Science · 2020
Abstract
Arousal labeling is one of the important methods in the diagnosis and treatment of sleep-related diseases, and are usually analyzed manually by doctors based on polysomnography (PSG) signals. In order to solve the problem of time-consuming and labor-intensive manual arousal analysis in sleep physiological signals, we propose an automatic arousal detection method using multi-model deep neural networks. Combining methods such as one-to-many formulation, LSTM, and network structure improvements, the performance of deep neural network models on clinical data set has been significantly improved, and multiple indicators have been improved (precision 86.7%, recall 86.0% and F1 86.3%). At the same time, the model parameters have been greatly streamlined, making them more efficient, laying a foundation for the application of automatic arousal detection methods on wearable sleep monitoring device signal analysis.