SleepNetZero: Zero-Burden Zero-Shot Reliable Sleep Staging With Neural Networks Based on Ballistocardiograms

TL;DR

This work introduces SleepNetZero, a zero-shot learning based approach for sleep staging that represents the first known reliable BCG-based sleep staging effort and marks a significant step towards in-home health monitoring.

Abstract

Sleep monitoring plays a crucial role in maintaining good health, with sleep staging serving as an essential metric in the monitoring process. Traditional methods, utilizing medical sensors like EEG and ECG, can be effective but often present challenges such as unnatural user experience, complex deployment, and high costs. Ballistocardiography~(BCG), a type of piezoelectric sensor signal, offers a non-invasive, user-friendly, and easily deployable alternative for long-term home monitoring. However, reliable BCG-based sleep staging is challenging due to the limited sleep monitoring data available for BCG. A restricted training dataset prevents the model from generalization across populations. Additionally, transferring to BCG faces difficulty ensuring model robustness when migrating from other data sources. To address these issues, we introduce SleepNetZero, a zero-shot learning based approach for sleep staging. To tackle the generalization challenge, we propose a series of BCG feature extraction methods that align BCG components with corresponding respiratory, cardiac, and movement channels in PSG. This allows models to be trained on large-scale PSG datasets that are diverse in population. For the migration challenge, we employ data augmentation techniques, significantly enhancing generalizability. We conducted extensive training and testing on large datasets~(12393 records from 9637 different subjects), achieving an accuracy of 0.803 and a Cohen's Kappa of 0.718. ZeroSleepNet was also deployed in real prototype~(monitoring pads) and tested in actual hospital settings~(265 users), demonstrating an accuracy of 0.697 and a Cohen's Kappa of 0.589. To the best of our knowledge, this work represents the first known reliable BCG-based sleep staging effort and marks a significant step towards in-home health monitoring.

Figures (7)

• Figure 1: The overview of the proposed framework. Training data from clinical datasets and test data from application requests are aligned through the component extraction module, which comprises three extractors. For training data, the extracted components are augmented by the generalization module, which consists of amplification and speed perturbation. Finally, these components are fed into a neural network for training and inference.
• Figure 2: The t-SNE visualization of the proposed data augmentation technique in the spectrum space. The original distributions of NSRR (blue) and BCG265 (red) data are distinct and separate. They cluster closer to each other (green, yellow) after data augmentation, illustrating that our data augmentation effectively eliminates the disparity between the two datasets.
• Figure 3: This is the architecture of the proposed three feature extractors, key components that extract high-dimensional features and fuse different modalities. The upper left part shows the overall architecture. The three components -- heartbeat, body movement, and breath -- are taken as inputs. Three ResNet-based feature extractors, denoted by ResFeat in the figure, produce high-dimensional representations for the three distinct groupings of components. The representations provided by the three extractors are concatenated together. The lower left part shows the detailed architecture of the ResFeat module, which is the main body of the feature extractor, denoted $\mathbf f_1, \mathbf f_2, \mathbf f_3$ in the main text. The right part shows the detailed architecture of the BasicBlock module, which is part of the ResFeat module.
• Figure 4: The prototype of the monitoring pad with the BCG sensor.
• Figure 5: The monitoring pad and the unencapsulated piezoelectric sensors. The sensors are already placed in their related encapsulation positions.