Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

SmartVital - Advanced Cardiac Health Monitoring

Presentation held at 1st European Conference in eHealth and Telemedicine in Cardiovascular Prevention and Rehabilitation, ETCPR 2013, June 7-8, 2013, Bern, Schweiz
: Noack, Aexander; Pietzsch, Marcus; Zimmerling, Martin; Poll, Rüdiger

presentation urn:nbn:de:0011-n-2721066 (266 KByte PDF)
MD5 Fingerprint: cde4be4afcd42e91018a6d6e44817f06
Created on: 18.12.2013

2013, 14 Folien
European Conference in eHealth and Telemedicine in Cardiovascular Prevention and Rehabilitation (ETCPR) <1, 2013, Bern>
Presentation, Electronic Publication
Fraunhofer IPMS ()

Fraunhofer IPMS is developing a ambulant holter monitoring device (SmartVital), which is capable of wireless data transmission. Wireless communication however, is very energy consuming. The operational hours of a device are therefore heavily influenced by the amount of transmitted data. Preliminary data analysis is one way to reduce data transmission, by determining which findings need to be transmitted and reducing the quantity of data to a minimum of significant values.
SmartVital is a M3-Cortex microprocessor based wearable ECG system, embedding highly optimized firmware with regard to energy consumption. The standard ECG signal is filtered (0,05Hz-40Hz) and AD-converted (12 bit / 250 Hz). Signal recordings are supplemented by tracking the activity of the patient via acceleration sensor values and air pressure measurements. However, to allow intelligent reporting device integrated algorithms are needed.
We introduce a real-time algorithm for ECG shape clustering. It works under computational constraints and is integrated at an early stage of the signal processing chain. Using a pseudo classification method it yields for normal QRS types 98.76 % sensitivity and 99.16 % positive prediction as well as 97.61% sensitivity and 99.64% positive prediction for ventricular ectopic beat morphologies on the MIT BIH Arrhythmia Database.
The algorithm is useful to separate diagnostically relevant signal morphologies. With low computational effort the beat identification quality exhibits similar results as existing methods described in literature. Combined with the already implemented activity state analysis this approach may identify normal cardiovascular or noisy signal episodes and prevent unnecessary data transmission.