direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Inhalt des Dokuments

Home-Based Sleep Apnea Screening

In recent years, a strong trend towards the development and use of unobtrusive medical instrumentation is evident. Besides being driven by the rapid development in the fields of sensors, integrated circuits and microprocessors, a higher integration of medical instrumentation naturally broadens their usability in a wider range of applications. Especially in the context of home-based monitoring, patient acceptance is a key element for acquiring usable, continuous data. Our goal in this research project is to develop algorithms and sensors to increase compliance, reduce complexity and ease the application of a home-based sleep apnea screening for pre-selection purposes.


Especially in the context of sleep monitoring, low patient obstruction is essential. With about 25% in adults, sleep disorders are amongst the most common medical conditions. The vast majority of sleep disorders are sleep disordered breathing issues which can have a severe impact on everyday life. Especially with respect to focus alertness in professional drivers, reduced sleep quality can lead to devastating consequences for the individual as well as society. As a consequence, the US passed a law in 2013 requiring the implementation of screening, testing or treatment of sleep disorders for professional drivers. A similar development is shown in the EU. Here, a directive was passed in 2014 requiring professional drivers at risk of sleep disorders to obtain a medical report. In both cases, the obstructive sleep apnea plays a central role.

Sleep disorders are usually diagnosed using polysomnography (PSG). Because of the large number of different recorded biosignals, PSG examinations are not only highly uncomfortable for the patient, they are also very cost intensive. In addition, obtaining an appointment already proves difficult, as there are not enough sleep laboratories available. Due to the previously mentioned regulations in both the US and the EU, an even higher demand for sleep monitoring can be expected.

Concept of a multi-functional sensor.

A possible solution to reduce the number of full PSGs is a preselection using home-based sleep monitoring. Currently there are different systems available which range from App-based concepts to somnographic systems which implement a stripped down version of the full PSG. The former relies on indirect measurements only, but allows therefore for a very high compliance. The latter, on the other hand, records biosignals directly, but is inherently limited in compliance due to a nasal sensor, finger clip and one or two respiration belts. In this project we seek to find the perfect balance between classification quality and comfort. For that purpose we are employing innovative combinations of multimodal sensors and fusion algorithms to approximate biosignals which otherwise could only be recorded in an obtrusive way.

Multi-Functional Sensor Systems

Multi-Functional Sensor.

The increasing integration and focus on acceptance comes at the cost of potentially using measurement sites which differ from gold standards. Therefore, evaluation signal quality and performance at non-standard measurement sites with respect to the gold standards is needed. By introducing multi-functional sensors, complex measurement setups can be further reduced in terms of sensor number.

Recently we examined different body positions for recording the surface ECG with inter-electrode distances down to 24 mm [1]. We analyzed the quality of the signals ECG signals with respect to standard Einthoven leads using original data recorded with our Robust Body Sensor Network (rBSN). In addition, we assessed the quality of derived respiration (DR) at different positions and combined both parameters to find sensor positions suitable for a multi-functional sensor.

Minimal ECG and derived respiratory parameters.

We are currently investigating and integrating additional sensors such as reflective PPGs, acoustical and impedance sensors and many more. The data retrieved from these sensors are then used to develop and evaluate classification and regression algorithms.

Multimodal Sensor Fusion

Neural Network.

An approach other than using more convenient or localized measurement sites is to extract parameters from other biosignals. By fusing multimodal sensors, not directly or only highly obstructively measureable biosignals can be obtained. Prominent examples are the blood pressure estimation using pulse transit time or the estimation of the respiratory signals without using sensor at the nose or the mouth.

Currently we are exploring different biosignal and parameter combinations in conjunction with classification and regression to approximate the respiratory flow and other only obtrusively obtainable biosignals. Besides classical approaches as Bayes classification we also use neural network based regression.

Features for classes inspiration and exspiration.

In the near future we will extend our research to incorporate a multitude of different biosignals in our framework. We will increase the number of independent signals recorded at a single measurement site and address problems such as classification of central and obstructive apnea, sleep onset detection and sleep duration assessment as well as sleep apnea related quality index estimation.

Related Work

[1] Klum, Michael, et al. "Minimally spaced electrode positions for multi-functional chest sensors: ECG and respiratory signal estimation.Current Directions in Biomedical Engineering 2.1 (2016): 695-699.

Zusatzinformationen / Extras


Schnellnavigation zur Seite über Nummerneingabe