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.

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.

Recent deep learning studies for sleep stage classification with polysomnography (PSG) data show two directions, either using 1-dimensional (1-D) raw PSG data or spectrogram images time-frequency domain. We propose a novel approach using images generated from time-signal display of a PSG dataset for 5 class sleep stage classification. The motivation of our approach is not only to imitate the way used by human sleep scoring experts but also to make use of various methods developed in image classification in Deep Learning, such as augmentation techniques, EfficientNet and LSTM. In addition an explainable AI technique such as Class Activation Map (CAM) can be employed for interpreting how a model makes a decision. We, also, work on “inconsistency” problems occurring among multiple institutions/hospitals where different capturing sensors are used and the labelling mismatch by human experts in different organizations. To solve the problem, we experiment three different approaches in the network design with data of two institutes and 5 sleep stage classification; (i) 5 class classification, (ii) 10-class classification and then post-processing to 5 classes (iii) 10-to-5 class classification. The 10-to-5 class classification is a network where information of two institutes are embedded inside the network. When information of multi institution is inside the network, the results show higher performance. Our experimental results show that all of three proposed methods based on time-signal images achieves higher accuracy performance compared to state-of-the-art models.

객체 추적(Object Tracking)은 최근 몇 년간 딥 러닝의 발전에 영향을 받아 급격하게 발전했다. 딥 러닝을 통해 학습된 객체 검출 시스템과 결합된 객체 추적 시스템들은 뛰어난 성능을 보여주지만 고성능의 연산 시스템을 필요로 한다. Simple Online and Realtime Tracking(SORT)는 다중 객체 추적 알고리즘으로 별도의 객체검출 시스템과 결합하여 이용된다. SORT는 적은 수의 객체를 추적하는 상황에서도 비교적 많은 연산량을 필요로 한다. 본 논문에서는 SORT를 유클리드 거리를 이용한 객체 추적 방법과 결합한 혼합형 추적 시스템을 제안하고 제안된 방법으로 상황에 따라 적응적으로 추적 방법을 바꾸는 객체 추적 알고리즘을 구현하였으며 기존의 SORT와 비교하여 효과적으로 연산량을 줄였음을 보인다.