Invention:
This invention is an end-to-end deep learning algorithm to detect cortical arousals during sleep using a one-night single lead electrocardiogram (ECG) signal. It is used to accurately detect arousal in home sleep tests, by reducing the false negative rate. The model combines both convolutional neural networks (CNN) and recurrent neural networks (RNN).
Background:
Sleep apnea is a side effect caused by excess weight and obesity, affecting more than a billion people worldwide. Sleep apnea is a ubiquitous sleep-related respiratory disease. Long term occurrences may lead to serious cardiovascular and neurological disease. Diagnosing and monitoring sleep apnea can happen in a lab or hospital with devices on the patient or at home. Home sleep testing is the preferred method for evaluation. Continuous Positive Airway Pressure (CPAP) is a highly recommended therapy for sleep apnea patients. It involves a machine that uses mild air pressure to keep airways open, helping an apneic person get sound sleep. However, despite its popularity, CPAP has disadvantages that are likely to negatively affect the sleep apnea treatment devices market growth. For example, CPAP machines are infamous for causing discomfort as patients find it difficult to sleep in specific positions to not to displace the mask on their face. Also, recording electroencephalogram (EEG) data is very inconvenient at home. There is a need for an improved method to detect cortical arousal to assist in the monitoring and diagnosis of sleep apnea with less noise and less intervention to home sleep testing patients.
Applications:
- Sleep apnea
- Deep learning ECG signal algorithm
- Recurrent neural networks
- Home sleep test
- Long-term in-home healthcare
- Emergency care in ambulatories
- Monitoring in intensive care units
Advantages:
- Ease of use
- Reduction of false negatives
- Improved efficiency
