Contactless apnea event detection using visible-thermal imaging

Apnea detection is a significant health concern due to its potential consequences, ranging from increased blood pressure to heart failure. Polysomnography is currently the gold standard for identifying apnea patterns during sleep. However, it requires trained personnel for analysis and is not suitable for long-term monitoring due to discomfort. To address these limitations, this paper proposes a contactless approach for apnea detection. The proposed approach utilizes visible and thermal imaging to remotely measure the breathing signal. This signal is then fed into deep learning models, including a 1-dimensional convolutional neural network (CNN), a long short-term memory (LSTM) network, and a hybrid model combining both. The effectiveness of these models is evaluated through comparative analysis. To evaluate the performance of the models, the authors define an apnea index to assess the presence of apnea in per second overlapped epochs. The validation of the contactless approach is evaluated by comparing the apnea detection results with those obtained from a contact-based breathing signal. The results demonstrate promising performance for each model. The mean absolute error values are reported as 0.6195 for CNN, 1.0177 for LSTM, and 1.3540 for CNN-LSTM. The Bland-Altman and correlation plot analyses demonstrate a high level of agreement between the contactless approach and the traditional contact-based method. Consequently, this approach might be useful for applications such as home-based patient monitoring, sleep studies, and neonatal apnea detection.