Application
Sleep stage classification algorithm and diagnostic platform.
Key Benefits
- Real-time sleep classification using a single electrocardiogram sensor.
- Eliminates the need for multiple head, eye, skin, and heart sensors with current diagnostics.
- Preliminary results using human data demonstrate high accuracy in identifying different sleep stages.
- Potential to enter the $500 million sleep diagnostic market.
Market Summary
Sleep disorders are a significant public health problem, affecting over 50 million in the United States and leading to over $94 billion in healthcare-associated costs. These conditions can lead to various health problems, including hypertension, heart disease, stroke, depression, diabetes, and chronic diseases. The gold standard for diagnosing sleep disorders is polysomnography (PSG), which involves attaching sensors to the patient's body to measure brain wave activity, eye movement, muscle tone, heart rhythm, and breathing. Unfortunately, PSG is uncomfortable and invasive to patients, often leading to misdiagnosis and treatment. Hence, there is a significant need to develop new technologies to accurately diagnose and treat sleep disorders.
Technical Summary
Emory researchers have developed a novel algorithm for the automatic classification of sleep stages by using a single lead electrocardiogram (ECG). The system processes ECG signals with a convolutional neural network (CNN) to identify characteristic features of each sleep stage. Thus far, the inventors have created a prototype of the system and trained it using a dataset containing 2829 30-second ECG signals from 16 human subjects annotated with sleep stages. The algorithm achieved an accuracy of 74% in classifying four sleep stages (wake, REM, NREM light, and NREM deep), 79% in classifying three sleep stages (wake, REM, and NREM; or wake/REM, NREM light, and NREM deep), and 83% in classifying two sleep stages (wake/REM vs NREM).
Developmental Stage
Early stages of development with results using human data.