Abnormal respiratory patterns classifier may contribute to large-scale screening of people infected with COVID-19 in an accurate and unobtrusive manner (2002.05534v2)

Abstract: Research significance: The extended version of this paper has been accepted by IEEE Internet of Things journal (DOI: 10.1109/JIOT.2020.2991456), please cite the journal version. During the epidemic prevention and control period, our study can be helpful in prognosis, diagnosis and screening for the patients infected with COVID-19 (the novel coronavirus) based on breathing characteristics. According to the latest clinical research, the respiratory pattern of COVID-19 is different from the respiratory patterns of flu and the common cold. One significant symptom that occurs in the COVID-19 is Tachypnea. People infected with COVID-19 have more rapid respiration. Our study can be utilized to distinguish various respiratory patterns and our device can be preliminarily put to practical use. Demo videos of this method working in situations of one subject and two subjects can be downloaded online. Research details: Accurate detection of the unexpected abnormal respiratory pattern of people in a remote and unobtrusive manner has great significance. In this work, we innovatively capitalize on depth camera and deep learning to achieve this goal. The challenges in this task are twofold: the amount of real-world data is not enough for training to get the deep model; and the intra-class variation of different types of respiratory patterns is large and the outer-class variation is small. In this paper, considering the characteristics of actual respiratory signals, a novel and efficient Respiratory Simulation Model (RSM) is first proposed to fill the gap between the large amount of training data and scarce real-world data. The proposed deep model and the modeling ideas have the great potential to be extended to large scale applications such as public places, sleep scenario, and office environment.

• The paper presents a novel deep learning classifier that distinguishes abnormal respiratory patterns for COVID-19 screening with 94.5% accuracy.
• It details a multi-stage methodology, combining a respiratory simulation model with real data from 20 subjects and depth camera recordings.
• The BI-AT-GRU model leverages bidirectional and attention mechanisms to outperform state-of-the-art classifiers for large-scale, unobtrusive monitoring.

Analysis of an Abnormal Respiratory Patterns Classifier for COVID-19 Screening

The research paper, "Abnormal Respiratory Patterns Classifier may contribute to large-scale screening of people infected with COVID-19 in an Accurate and Unobtrusive manner," proposes a novel computational approach for distinguishing between several respiratory patterns, with a particular focus on applications in COVID-19 detection. Utilizing the distinct respiratory characteristics associated with COVID-19, such as Tachypnea, the paper leverages depth cameras and deep learning techniques to develop a non-contact screening tool. This tool may facilitate large-scale, unobtrusive respiratory monitoring in various environments, potentially aiding in the early identification and monitoring of COVID-19 cases.

Methodological Framework

The paper outlines a multi-stage approach to the development and validation of a respiratory pattern classifier using a Gated Recurrent Unit (GRU) neural network enhanced with bidirectional and attentional mechanisms (BI-AT-GRU). The methodological framework comprises the following key steps:

1. Respiratory Simulation Model (RSM): Given the scarcity of real-world data necessary for effective model training, a Respiratory Simulation Model was developed to generate synthetic yet realistic training data. This model simulates respiratory signals using a sine wave approximation that accounts for environmental noise and motion artifacts, thereby producing diverse training datasets reflective of real-world conditions.
2. Data Acquisition via Depth Camera: Real-world respiratory data from 20 subjects were collected using Kinect v2 depth cameras. Participants were trained to simulate six distinct respiratory patterns, including both normal (Eupnea) and abnormal patterns (Bradypnea, Tachypnea, Biots, Cheyne-Stokes, and Central-Apnea).
3. Model Definition and Training: The BI-AT-GRU model was trained using the synthetic data from the RSM. This architecture capitalizes on bidirectional processing to account for sequence information and incorporates attention mechanisms to focus on pivotal waveform regions associated with specific respiratory patterns.
4. Validation and Comparative Analysis: The model was rigorously validated using 605 real-world data samples. Performance metrics indicated that the BI-AT-GRU model achieved superior classification accuracy and outperformed existing state-of-the-art models, delivering an accuracy of 94.5%, precision of 94.4%, recall of 95.1%, and an F1 score of 94.8%.

Implications and Future Directions

The paper presents significant implications for both clinical practice and broader public health initiatives. The ability to accurately classify respiratory patterns using a non-contact depth camera offers a viable solution for remote monitoring in diverse settings, such as public spaces, workplaces, and hospitals. This technology could prove particularly beneficial in scenarios where traditional contact-based monitoring methods are impractical, thereby enhancing the accessibility and scalability of respiratory monitoring during pandemic situations.

Moreover, the introduction of the RSM provides a foundation for generating training data in other contexts where real-world data are limited. Given the promising results demonstrated by BI-AT-GRU, the architecture may be adapted and optimized for use in other domains requiring sequential data classification, extending its utility beyond respiratory pattern analysis.

While the paper provides robust initial findings, future research could focus on enhancing the system's robustness in the presence of varying environmental conditions and expanding its applicability across diverse demographic groups. Additionally, the integration of more advanced image processing techniques and AI-driven data analytics could improve the sensitivity and specificity of respiratory anomaly detection further.

Overall, the research highlights a significant step forward in the application of AI and deep learning in biomedical engineering, providing a template for future innovations in unobtrusive physiological monitoring technologies.