Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Sleep Stages Algorithm Based On Wearable And Nearable Sensors Fusion
This project aims to combine wearable and contactless devices in a sensor fusion fashion to measure sleep in a real hospital setting with hundreds of patients suffering from sleep disorders.
Several sleep/wake classification algorithms for wearable devices have been suggested over the last decades, and they are typically based on actigraphy derived from an accelerometer. Several findings suggest that only using movement signals leads to the main limitation of current algorithms: the incorrect classification and overestimation of low-activity tasks such as sleep. Recent studies in the field leverage a combination of photoplethysmography (PPG) and accelerometer signals for sleep/wake detection. However, there is still a lack in the literature of algorithms capable of classifying different sleep stages, including REM, light, and deep sleep, using a combination of physiological data such as activity, heart rate, electrodermal activity (EDA), and skin temperature, especially in patients with sleep disorders.
Several sleep/wake classification algorithms for wearable devices have been suggested over the last decades, and they are typically based on actigraphy derived from an accelerometer. Several findings suggest that only using movement signals leads to the main limitation of current algorithms: the incorrect classification and overestimation of low-activity tasks such as sleep. Recent studies in the field leverage a combination of photoplethysmography (PPG) and accelerometer signals for sleep/wake detection. However, there is still a lack in the literature of algorithms capable of classifying different sleep stages, including REM, light, and deep sleep, using a combination of physiological data such as activity, heart rate, electrodermal activity (EDA), and skin temperature, especially in patients with sleep disorders.
This project aims to develop an algorithm to classify four sleep stages: wake, REM, light, and deep sleep, using in the first phase a combination of activity and heart rate data from radar, a sensing mattress, and a wristband in a sensor fusion fashion. In the second phase, it is possible to add EDA and skin temperature data from a wristband and compare the algorithms' performances. The student will be able to use data collected in a real clinical setting with patients with different sleep disorders. Polysomnography (PSG) will be used as the ground truth data.
This project aims to develop an algorithm to classify four sleep stages: wake, REM, light, and deep sleep, using in the first phase a combination of activity and heart rate data from radar, a sensing mattress, and a wristband in a sensor fusion fashion. In the second phase, it is possible to add EDA and skin temperature data from a wristband and compare the algorithms' performances. The student will be able to use data collected in a real clinical setting with patients with different sleep disorders. Polysomnography (PSG) will be used as the ground truth data.
Work with a real dataset collected in an unsupervised setting, a sleep clinical hospital with patients diagnosed with different sleep disorders.
Contribute to developing an algorithm to find the most accurate, and unobtrusive way to measure sleep.
The project is led and supervised by experts in data analysis, machine learning, and experts from the clinical field. You can benefit from mentoring in both areas.
Work with a real dataset collected in an unsupervised setting, a sleep clinical hospital with patients diagnosed with different sleep disorders.
Contribute to developing an algorithm to find the most accurate, and unobtrusive way to measure sleep.
The project is led and supervised by experts in data analysis, machine learning, and experts from the clinical field. You can benefit from mentoring in both areas.
- Strong programming skills in Python.
- Strong background in machine learning.
- Structured and reliable working style.
- Ability to work independently on a challenging topic.
- Proven records on some of the following: experience in large data sets, longitudinal data analysis, and sparse feature selection.
- Strong programming skills in Python. - Strong background in machine learning. - Structured and reliable working style. - Ability to work independently on a challenging topic. - Proven records on some of the following: experience in large data sets, longitudinal data analysis, and sparse feature selection.
If interested, please contact me via email with your CV and transcripts attached: Oriella Gnarra oriella.gnarra@hest.ethz.ch
If interested, please contact me via email with your CV and transcripts attached: Oriella Gnarra oriella.gnarra@hest.ethz.ch