Project Description

Nursing support for people with epilepsy through innovative ear sensor technology (EPItect)

Within the EPItect research project, a non-invasive sensor system is being developed to detect epileptic seizures based on relevant biosignal patterns. An automatic and reliable seizure detection is of central importance for the correct adjustment of the patient’s medication and is a great help for the patient’s nursing environment.

Background

According to the World Health Organization (WHO), there are 50 million people worldwide with epilepsy. All age groups and sexes are affected. By disrupting brain signals, epilepsy causes seizures of varying severity. Many people with epilepsy are able to live a largely normal life and perform the same activities as others. However, some epileptics may be severely affected in their lives because of frequent seizures.

Motivation and challenge

Epileptic seizures are sometimes associated with considerable risks and can lead to accidents with serious injuries. For this reason it is important to recognise epileptic seizures in good time in order to take appropriate safety precautions. In addition, automatic and reliable seizure detection is essential for effective medication of patients. In practice, EEG (electroencephalography) is most frequently used for seizure detection. By measuring and monitoring brain waves, seizures can be predicted. However, the fact that this type of monitoring can only be carried out as an in-patient in hospital is a major limitation. Up to now, there is no technical solution for seizure detection in the everyday life of patients. Seizures are independently documented by the patients themselves by means of a hand log. However, these hand logs do not offer sufficient reliability due to the high susceptibility to errors. According to a study1, only 44.5% of seizures are registered in this way.

Goals

With the help of the earconnect™ technology from cosinuss°, the aim is to revolutionize seizure detection for epilepsy patients in everyday life:

  • Automatic seizure detection in everyday life
  • High sensitivity and specificity
  • Non-invasive in-ear sensor system
  • Mobile data exchange for selected persons
  • Suitability for everyday use and high acceptance by those concerned
  • Compliance with data protection regulations
  • Alarm service and emergency call system for critical attacks
  • Improving the safety, self-determination and quality of life of epilepsy patients

Why cosinuss°?

The earconnect™ technology from cosinuss° is ideal for detecting epileptic seizures. The in-ear sensor technology provides a mobile, continuous and precise measurement of the heart rate. It can also detect blood pressure fluctuations specific to seizures. Furthermore, the cosinuss° In-Ear Sensors are able to measure temperature and movements with high precision.

For the patient himself, measurement in the ear offers several advantages for seizure detection:

  • The application of the sensors is simple
  • No cabling is necessary
  • The sensor system does not interfere with everyday tasks
  • The sensor is protected from seizure-associated extremity movements

Test procedure

Information Description
Type of study Clinical trial
Used devices cosinuss° One
Reference / Technical validity EEG video monitoring (current gold standard for epileptic seizures), ECG (current gold standard for heart rate measurement)
Time frame March 2016 to February 2019
Location of measurement Inpatient (University Clinic for Epileptology Bonn)
Number of test persons 182
Test person statistics
  • Epilepsy patients
  • 43.2 +/- 17.7 years
  • Female: 54%, Male: 46
Protocol
  • The data collection was carried out by the medical staff.
  • Sensors were changed every six hours.
  • Long-term video EEG (1-13 days), ECG, pulse oximetry
  • Continuous measurement.
Collected data heart rate, RR intervals, temperature, PPG, 3D acceleration data
Data transmission cosinuss° One > cosinuss° LapApp > cosinuss° LabServer
Data analysis More than 20,000 hours of data collected using the sensors of almost 200 patients are evaluated by a learning algorithm to classify the seizures. Various types of machine learning and deep learning are validated and, if necessary, improved. The algorithms are fed with data that are processed using cosinuss’ proprietary pre-processing methods.

Conclusions based on the data

The results so far are promising, but also show a need for optimisation. Some types of epileptic seizures are difficult to detect. This makes it difficult to find suitable pre-processing methods. In addition, not enough data are yet available for all types of seizures. An effective training of the algorithms can only be carried out on the basis of sufficient available data for specific types of seizures.

In addition, due to the limited sample of almost 200 people, the accuracy of seizure detection is 40%.

Outlook

Based on the results of the EPItect research project, we are striving for a significant improvement in epileptic seizure detection in the follow-up research project MOND. For this purpose, additional vital parameters and biosignals are used

Possible areas of application

Telemonitoring of epileptic seizures and other diseases in everyday life.

About this study

Network Coordinator:

University Clinic for Epileptology Bonn (Uniklinik für Epileptologie Bonn)

Partner:

University Clinic for Epileptology Bonn, Cosinuss GmbH, Fraunhofer Institute for Software and Systems Engineering ISST, Clinic for Neuropaediatrics at the University of Kiel, North German Epilepsy Center in Schwentinental-Raisdorf

Associated partners:

University of Applied Sciences Bochum, Epilepsy Federation of Parents’ Association e.V. Wuppertal, State Association for Epilepsy Self-Help North Rhine-Westphalia e.V.

Funding:

Federal Ministry of Education and Research in the programme “Care innovations to support informal and professional carers”

Federal Ministry of Education and Research

Project Management:

VDI/VDE Innovation + Technik GmbH; Volumen: 2,3 Mio. € (of which 85% funded by the
Bundesministerium für Bildung und Forschung)

VDI/VDE Innovation + Technik GmbH

References

  1. Hoppe, C., Poepel, A., & Elger, C. E. (2007). Accuracy of patient seizure counts.