Recently, deep learning for automated sleep stage classification has been introduced with promising results. However, as many challenges impede their routine application, automatic sleep scoring algorithms are not widely used. Typically, polysomnography (PSG) uses multiple channels for higher accuracy; however, the disadvantages include a requirement for a patient to stay one or more nights in the lab wearing uncomfortable sensors and wires. To avoid the inconvenience caused by the multiple channels, we aimed to develop a deep learning model for use in clinical decision support systems (CDSSs) and combined convolutional neural networks and a transformer for the supervised learning of three classes of sleep stages only with single-channel EEG data (C4-M1). The data for training, validation, and test were derived from 1590, 341, and 343 polysomnography recordings, respectively. The developed model yielded an overall accuracy of 91.4%, comparable with that of human experts. Based on the severity of obstructive sleep apnea, the model’s accuracy was 94.3%, 91.9%, 91.9%, and 90.6% in normal, mild, moderate, and severe cases, respectively. Our deep learning model enables accurate and rapid delineation of three-class sleep staging and could be useful as a CDSS for application in real-world clinical practice.

Study Objectives: Polysomnography (PSG) scoring is labor-intensive, subjective, and often ambiguous. Recently several deep learning (DL) models for automated sleep scoring have been developed, they are tied to a fixed amount of input channels and resolution. In this study, we constructed a standardized image-based PSG dataset in order to overcome the heterogeneity of raw signal data obtained from various PSG devices and various sleep laboratory environments. Methods: All individually exported European data format files containing raw signals were converted into images with an annotation file, which contained the demographics, diagnoses, and sleep statistics. An image-based DL model for automatic sleep staging was developed, compared with a signal-based model, and validated in an external dataset Results: We constructed 10253 image-based PSG datasets using a standardized format. Among these, 7745 diagnostic PSG data were used to develop our DL model. The DL model using the image dataset showed similar performance to the signal-based dataset for the same subject. The overall DL accuracy was greater than 80%, even with severe obstructive sleep apnea. Moreover, for the first time, we showed explainable DL in the field of sleep medicine as visualized key inference regions using Eigen-class activation maps. Furthermore, when a DL model for sleep scoring performs external validation, we achieved a relatively good performance. Conclusions: Our main contribution demonstrates the availability of a standardized image-based dataset, and highlights that changing the data sampling rate or number of sensors may not require retraining, although performance decreases slightly as the number of sensors decreases.

Recent deep learning models succeed in achieving high accuracy and fast inference time, but they require high-performance computing resources because they have a large number of parameters. However, not all systems have high-performance hardware. Sometimes, a deep learning model needs to be run on edge devices such as IoT devices or smartphones. On edge devices, however, limited computing resources are available and the amount of computation must be reduced to launch the deep learning models. Pruning is one of the well-known approaches for deriving light-weight models by eliminating weights, channels or filters. In this work, we propose “zero-keep filter pruning” for energy-efficient deep neural networks. The proposed method maximizes the number of zero elements in filters by replacing small values with zero and pruning the filter that has the lowest number of zeros. In the conventional approach, the filters that have the highest number of zeros are generally pruned. As a result, through this zero-keep filter pruning, we can have the filters that have many zeros in a model. We compared the results of the proposed method with the random filter pruning and proved that our method shows better performance with many fewer non-zero elements with a marginal drop in accuracy. Finally, we discuss a possible multiplier architecture, zero-skip multiplier circuit, which skips the multiplications with zero to accelerate and reduce energy consumption.