|
 
 |
| ORIGINAL ARTICLE |
|
| Year : 2013 | Volume
: 3
| Issue : 2 | Page : 63-68 |
|
Detection of Epileptic Seizure Using Wireless Sensor Networks
Golshan Taheri Borujeny, Mehran Yazdi, Alireza Keshavarz-Haddad, Arash Rafie Borujeny
Department of Communication and Electronic Engineering, School of Electrical and Computer Engineering, Shiraz University, Shiraz, Iran
| Date of Submission | 18-Nov-2012 |
| Date of Acceptance | 25-Mar-2013 |
| Date of Web Publication | 20-Sep-2019 |
Correspondence Address:
 Source of Support: None, Conflict of Interest: None  | 26 |
DOI: 10.4103/2228-7477.114373
The monitoring of epileptic seizures is mainly done by means of electroencephalogram (EEG) monitoring. Although this method is accurate, it is not comfortable for the patient as the EEG-electrodes have to be attached to the scalp which hampers the patient's movement. This makes long-term home monitoring not feasible. In this paper, the aim is to propose a seizure detection system based on accelerometry for the detection of epileptic seizure. The used sensors are wireless, which can improve quality of life for the patients. In this system, three 2D accelerometer sensors are positioned on the right arm, left arm, and left thigh of an epileptic patient. Datasets from three patients suffering from severe epilepsy are used in this paper for the development of an automatic detection algorithm. This monitoring system is based on Wireless Sensor Networks and can determine the location of the patient when a seizure is detected and then send an alarm to hospital staff or the patient's relatives. Our wireless sensor nodes are MICAz Motes developed by Crossbow Technology. The proposed system can be used for patients living in a clinical environment or at their home, where they do only their daily routines. The analysis of the recorded data is done by an Artificial Neural Network and K Nearest-Neighbor to recognize seizure movements from normal movements. The results show that K Nearest Neighbor performs better than Artificial Neural Network for detecting these seizures. The results also show that if at least 50% of the signal consists of seizure samples, we can detect the seizure accurately. In addition, there is no need for training the algorithm for each new patient. Keywords: 2D accelerometer, epilepsy seizure detection, K Nearest-Neighbor, neural network, wireless sensor network
How to cite this article:
Borujeny GT, Yazdi M, Keshavarz-Haddad A, Borujeny AR. Detection of Epileptic Seizure Using Wireless Sensor Networks. J Med Signals Sens 2013;3:63-8 |
How to cite this URL:
Borujeny GT, Yazdi M, Keshavarz-Haddad A, Borujeny AR. Detection of Epileptic Seizure Using Wireless Sensor Networks. J Med Signals Sens [serial online] 2013 [cited 2023 Mar 29];3:63-8. Available from: https://www.jmssjournal.net/text.asp?2013/3/2/63/114373 |
| Introduction | |  |
Epilepsy is one of the most common neurological disorders, affecting almost 60 million people all over the world. Most of the affected people can be treated successfully with drug therapy (67%) or neurosurgical procedures (7%-8%). Nevertheless 25% of the affected people cannot be treated by any available therapy. [1] For refractory patients who continue to have frequent seizures, it has been shown that intensive monitoring with electroencephalogram (EEG) and video over a long period, contributes to the management of daily care and the adjustment of drug therapy. [2] The long-term monitoring with EEG and video can be very unpleasant for patients, and analyzing large amounts of EEG/video-data is very labor intensive for medical personnel. Furthermore, this method cannot yet be applied in real-time procedures.
All the above-mentioned factors have made it necessary to look for sensors that are patient friendly and can be used for a reliable automatic detection of epileptic seizures. One of these sensors is the accelerometer. Accelerometers are used in many medical research areas for activity recognition. [3],[4],[5] For instance, in Parkinson's disease, studies aim at distinguishing pathological (periods of hypokinesia, bradykinesia and dyskinesia) and normal movements. [6],[7]
Here, we focus on motor signs since epileptic seizures are often accompanied by motor signs. With accelerometry (ACM), only seizures that express themselves in movements or disturb normal movement patterns can be detected. The results in reference [8] demonstrate that 3D ACM is a valuable sensing method for seizure detection. The authors in reference [9] propose an algorithm based on HMM for seizure detection. As in reference [9] , in reference [10] the patient movements are modeled with Hidden Markov Models and Bayesian analysis of the signal is performed. Algorithm of seizure detection in reference [9] is without adaptation to patient but this problem is overcome in reference [10] . About nocturnal epileptic seizure, reference [11] focuses on the distinction between seizure moves and nocturnal moves. Sensors are therefore attached on a patient and the authors propose to detect period with motor activities.
Long-term home monitoring can provide the neurologist an objective measure of the number of seizures that a patient can have during the day. Also in some of the heavy epileptic attacks the patient needs medical care after or during the seizure. Human body movement can be monitored through a wireless network composed of inertial sensors. Reference [12] presents the development of Wagyromag (Wireless Accelerometer, GYROscope, and MAGnetometer), a wireless Inertial Measurement Unit (IMU) composed of a triaxial accelerometer, gyroscope and magnetometer. The authors describe a seizure detection system based on Wireless Sensor Network (WSN) that can determine the location of the patient when a seizure is detected and sends an alarm to hospital staff or the patient's relatives.
Proposed System
In this paper, we introduce a system based on WSN that provides a continuous monitoring without limiting the freedom and privacy of the patients. The main goal is to distinguish between data with and without seizure movement.
The general constraints and characteristics of the system are listed as follows:
- Flexibility: All external wirings can be removed to allow the subject under test move without restrictions
- Ease of use: The network protocol allows the sensor network to initialize itself in a highly ad-hoc, self-organizing manner
- Reliability: While data reliability is always important, it becomes a critical requirement for many applications, for example, in medical monitoring
- Biaxial acceleration measurement
- Rechargeable batteries
- Low power consumption
- Comfortability: The device has small size and low weight, and is able to be attached to different parts of patient's body.
There are two kinds of nodes in the network:
- Mobile sensor nodes which are placed on the body of patients
- Static nodes which are sited on fixed specific locations at the building.
The static nodes transport the collected data from mobile sensor nodes to a base- station. The base- station sends data to computer server through a USB cable. Recorded data will be processed and when a seizure is detected, static nodes can determine approximately the location of the patient and sends an alarm to hospital staff or the patient's relatives.
The analysis of the recorded data is based on artificial neural network (ANN) and K Nearest Neighbor (KNN) to recognize seizure movements from normal movements. [Figure 1] shows the proposed system for epileptic seizure detection.
| Seizure Detection Method | | |
Data Collecting
Datasets from patients suffering from heavy epilepsy were used for the development of an automatic detection algorithm. In this system, three 2D accelerometer sensors were positioned on the right arm, left arm and left thigh of epileptic patients. Datasets were acquired from three patients suffering from severe epilepsy. The datasets of the epileptic patients were recorded during the day. We recorded 20 epileptic seizures.
Patients were asked to perform a sequence of everyday normal activities but were not told specially how to do them. Normal activities that we recorded included static activities such as reading, working with computer, brushing of teeth, and lying and dynamic activities such as walking. The sampling frequency of the accelerometer is 3 Hz. [Figure 2] shows the pure output of accelerometers when sampling frequency is 3 Hz. It shows at first lying and then seizure signal. In this Figure the seizure has begun from 180 samples. Acceleration has been measured based on gravity (g=9.8m/s2 )
For analyzing and detecting seizures from this huge data sequence, the best way is cutting the acceleration sequences into many overlapping windows (segments) of the same length. For our data, the size of this window is considered 50 samples and it is repeated for every 25 samples. Since the sampling frequency is 3 Hz, we cut the data sequence every 9 seconds and analyzed this window of the ACM data to detect seizure. [Figure 3] shows the acceleration data and overlapping window that located the signal.
Preprocessing
The output of an accelerometer attached to the human body consists of different components:
- Noise from sensor and measurement system
- Noise sources from the environment: (a) accelerations produced by external sources like vehicles; (b) accelerations due to bumping of the sensor or the body against other objects
- Noise sources from the body: (a) Muscle tremor; (b) Heart; (c) Respiration; (d) Blood flow
- Gravitational acceleration
- Acceleration due to movements of the body
In comparison to body movements, the noise from the sensor and measurement system can be neglected. All data used in this study were recorded while the patients were in their living environment, thus there were no accelerations produced by external sources.
When there is no movement, physiological perturbations, like respiration and heart rate and gravitational acceleration are visible in the signal. A preprocessing step is executed on the raw data for deleting these perturbations. To do that, we use a moving average filter. If the received signal is denoted as X(k), the filtered signal is given by following equation:[Figure 4]
Network Topology
There are several architectures that can be used to implement WSN applications, including star, mesh, and star-mesh hybrid. Each topology presents its own set of challenges, advantages, and disadvantages. The topology refers to the configuration of the hardware components and how the data are transmitted through that configuration.
XMesh is a full featured multi-hop, ad-hoc, mesh networking protocol developed by Crossbow, [15] for wireless network. An XMesh network consists of nodes that wirelessly communicate to each other and are capable of hopping radio messages to a base station where they are passed to a central server. The hopping effectively extends radio communication range and reduces the power required to transmit messages. By hopping data in this way, XMesh can provide two critical benefits: improved radio coverage and improved reliability. Two nodes do not need to be within direct radio range of each other to communicate. Xmesh provides to support both Zigbee standards (802.15.4) and advanced mesh networking. XMesh provides a TrueMesh networking service that is both self-organizing and self-healing. In [Figure 5] is presented the diagram for XMesh network.
We have implemented this WSN in a three-floor building that has 16 rooms in each floor.
Hardware Requirement
We have implemented this system with our wireless sensor nodes, Motes, developed by Crossbow Technology. The device is built upon the IEEE 802.15.4 standard and has an 8-bit Atmel ATmega microcontroller. It uses the Chipcon CC2420, ZigBee ready radio frequency transceiver designed for low-power and low-voltage wireless communication in the 2.4 GHz unlicensed ISM band. [15]
For the purpose of seizure detection, we use MTS310 sensor board. MTS310 has a variety of sensing modalities including an accelerometer two-axis device. [Figure 6] shows an MICAz Mote and an MTS310 sensor board. [16]
Locating the Patient
As mentioned above, the static nodes have an interface role between mobile sensor nodes and base-station. Also, location of each mobile node can be determined by the closest static node. So the location of static nodes must be chosen such that at any time, a mobile node connects to one and only one static node, since each mobile node is localized by the nearest static node. Although a large radio range for the nodes increases the power consumption and decreases the accuracy for localizing the mobile nodes, it helps to cover a building with fewer static nodes. So, we should set a trade-off between the number of nodes and radio range for them.
We set the node's power on 316 μW. In this power, the radio range of nodes is 30 m in a line of sight area and we can cover a building using eight static nodes.
| Experimental Results | | |
The proposed system in this research can be used for patients living in a clinical environment or at their home, where they do only their daily routine. We could detect seizures from normal activities such as walking. The analysis of the recorded data is based on the ANN and KNN to recognize seizure movement from normal movement. When we classified datasets of the three patients by ANN to detect epileptic seizures, we obtained the results shown in [Table 1].
As can be seen in the table, three seizures have not been detected and we have obtained three false alarms. Sensibility in this system is 85%.
Then we used KNN to classify the datasets and detect seizures. The results of using this classifier with k = {1, 3, 5} are shown in [Table 2].
[Table 2] shows that for k = 1 two seizures have not been detected and three normal activities have been detected as seizures. For k = 3 we have only two false alarms. But for k = 5 no seizure has been missed by the system and we have no false alarm. So KNN algorithm has produced better results than ANN.
| Discussion and Conclusion | | |
One of the most interesting applications of WSN is health monitoring. In this paper we introduced a monitoring system based on WSN for detection of epilepsy seizures. This system can be used for patients living in a clinical environment or at their home, where they do only their daily routine. Our system can determine the location of the patient when a seizure is detected and sends an alarm to hospital staff or the patient's relatives. Since our sensors are wireless, the subject under test can move without restrictions. In addition, because of the small size of the sensor nodes, they are wearable for patients. Our experimental results showed that 5 Nearest Neighbors provides better results than the ANN classifier. Furthermore, there is no need for training the algorithm for each new patient.
| References | | |
| 1. | Witte H, Iasemidis LD, Litt B. Special issue on epileptic seizure prediction. IEEE Trans Biomed Eng 2005;50:537-9.  |
| 2. | Binnie CD, Aarts JH, Van Bentum-De Boer PT, Wisman T. Monitoring at the institute for epilepsy fight in Bosch..Electroencephalogr and Clin Neurophysiol Suppl 1985;37:341-55. |
| 3. | Mathie MJ, Celler BG, Lovell NH, Coster AC. Classification of basic daily movements using a triaxial accelerometer. Med Biol Eng Comput 2004;42:679-87. |
| 4. | Veltink P, Bussmann HB, de Vries W, Martens WL, Van Lummel RC. Detection of static and dynamic activities using uniaxial accelerometers. IEEE Trans Rehabil Eng 1996;4:375-85. |
| 5. | Najafi B, Aminian K, Paraschiv-Ionescu A, Loew F, Bula C, Robert P. Ambulatory system for human motion analysis using a kinematic sensor: Monitoring of daily physical activity in the elderly. IEEE Trans on Biomed Eng 2003;50:711-23. |
| 6. | Dunnewold RJ, Hoff JI, Pelt HC, Fredrikze PQ, Wagemans EA, Hilten BJ. Ambulatory quantitative assessment of body position, bradykinesia, and hypokinesia in parkinson's disease. J Clin Neurophysiol 1998;15:235-42. |
| 7. | Thielgen T, Foerster F, Fuchs G, Hornig A, Fahrenberg J. Tremor in parkinson's disease: 24-hr monitoring with calibrated accelerometry. Electromyogr Clin Neurophysiol 2004;44:137-46. |
| 8. | Nijsen TM, Arends JB, Griep PA, Cluitmans PJ. The potential value of three-dimensional accelerometry for detection of motor seizures in severe epilepsy. Epilepsy and Behav 2005;7:74- 84. |
| 9. | Jallon P, Bonnet S, Antonakios M, Guillemaud R. Detection system of motor epileptic seizures through motion analysis with 3D accelerometers. Conf Proc IEEE Eng Med Biol Soc 2009; p. 2466-9. |
| 10. | Jallon P. A Bayesian approach for epileptic seizures detection with 3D accelerometers sensors. Conf IEEE Eng Med Biol Soc 2010;2010:6325-8. |
| 11. | Nijsen TM, Cluitmans PJ, Arends JB, Griep PA. Detection of subtle nocturnal motor activity from 3d accelerometry recordings in epilepsy patients. IEEE Trans Biomed Eng 2007;54:2073- 81. |
| 12. | Olivares A, Olivares G, Mula F, Górriz JM, Ramírez J. Wagyromag: Wireless sensor network for monitoring and processing human body movement in healthcare applications. J Sys Arch 2011;57:905-15. |
| 13. | Bao L, Intille SS. Activity Recognition from User-Annotated Acceleration Data. New York: Springer-Verlag Berlin Heidelberg 2004; p. 1-17. |
| 14. | Looney CG. Pattern recognition using neural networks, theory and algorithms for engineers and scientists. Oxford: Oxford University Press; 1997. |
| 15. | Available from: http://www.xbow.comIProducts/productdetails.aspx? sid=l64 [Last accessed on 2012 Nov 03]. |
| 16. | Available from: http://www.tinyos.millennium.berkeley.edul [Last accessed on 2012 Nov 03]. |
| Authors | | |
 Golshan Taheri Borujeny was born in 1985. She received the B.Sc. degree in Biomedical engineering from the Isfahan University, Isfahan and the M.Sc. degree in communication systems from the Department of communication and electronic engineering, Shiraz University, Shiraz, Iran, in 2007 and 2013, respectively. Her research interests include image processing, biomedical signal processing and movement analysis.
E-mail: [email protected]
 Mehran Yazdi received his B.Sc. degree in Digital Communication Systems from the Department of Electrical Engineering, Shiraz University, in 1992 and M.Sc. and Ph.D. degrees in Digital Vision and Image Processing from the Department of Electrical Engineering, Laval University, Canada, in 1996 and 2003 respectively. He is currently Associate Professor at the Department of Communication and Electronic Engineering, Shiraz University, Iran. He conducted several projects in the area of hyperspectral image compression and denoising, CT metal artifact reduction and video compression. His major research interests are in the field of Image/video processing, remote sensing, Multidimentional signal processing and medical image analysis.
E-mail: [email protected]
Alireza Keshavarz‑Haddad received his B.Eng. degree in 2001 from the Department of Electrical Engineering of Sharif University, Tehran, Iran, and his M.S. and Ph.D. degrees in 2003 and 2007 from the Department Electrical and Computer Engineering of Rice University, Houston, Texas. Since 2008, he has been with the School of Electrical and Computer Engineering at Shiraz University, Shiraz, Iran, where he is currently an Assistant Professor. His research interests are in Wireless Ad Hoc and Sensor networks, Network Coding, and Network Security.
E-mail: [email protected]
 Arash Rafie Borujeny was born in 1985. He received the B.Sc. degree in electrical engineering from the Shahid Rajaee Teacher training University, Tehran, Iran, in 2008. He is currently pursuing the M.Sc. degree in the Department of communication and electronic engineering, Shiraz University, Shiraz, Iran. His research interests are pattern recognition, neural network and numerical methods.
E-mail: [email protected]
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6]
[Table 1], [Table 2]
| This article has been cited by | | 1 |
Assessing the Feasibility of Acoustic Based Seizure Detection |
|
| Xuen Hoong Kok, Syed Anas Imtiaz, Esther Rodriguez-Villegas | | IEEE Transactions on Biomedical Engineering. 2022; 69(7): 2379 | | [Pubmed] | [DOI] | | | 2 |
A Two-Layer Ensemble Method for Detecting Epileptic Seizures Using a Self-Annotation Bracelet With Motor Sensors |
|
| Chunjiao Dong, Tianchun Ye, Xi Long, Ronald M. Aarts, Johannes P. van Dijk, Chunheng Shang, Xiwen Liao, Wei Chen, Wanlin Lai, Lei Chen, Yunfeng Wang | | IEEE Transactions on Instrumentation and Measurement. 2022; 71: 1 | | [Pubmed] | [DOI] | | | 3 |
An Intracortical Implantable Brain-Computer Interface for Telemetric Real-Time Recording and Manipulation of Neuronal Circuits for Closed-Loop Intervention |
|
| Hamed Zaer,Ashlesha Deshmukh,Dariusz Orlowski,Wei Fan,Pierre-Hugues Prouvot,Andreas Nørgaard Glud,Morten Bjørn Jensen,Esben Schjødt Worm,Slávka Lukacova,Trine Werenberg Mikkelsen,Lise Moberg Fitting,John R. Adler,M. Bret Schneider,Martin Snejbjerg Jensen,Quanhai Fu,Vinson Go,James Morizio,Jens Christian Hedemann Sørensen,Albrecht Stroh | | Frontiers in Human Neuroscience. 2021; 15 | | [Pubmed] | [DOI] | | | 4 |
Assessing the Feasibility of Detecting Epileptic Seizures using Non-Cerebral Sensor Data |
|
| Alexandra Hamlin,Erik Kobylarz,James H. Lever,Susan Taylor,Laura Ray | | Computers in Biology and Medicine. 2021; : 104232 | | [Pubmed] | [DOI] | | | 5 |
Non-invasive wearable seizure detection using long–short-term memory networks with transfer learning |
|
| Mona Nasseri,Tal Pal Attia,Boney Joseph,Nicholas M Gregg,Ewan S Nurse,Pedro F Viana,Andreas Schulze-Bonhage,Matthias Dümpelmann,Gregory Worrell,Dean R Freestone,Mark P Richardson,Benjamin H Brinkmann | | Journal of Neural Engineering. 2021; 18(5): 056017 | | [Pubmed] | [DOI] | | | 6 |
Telemedicine for the Care of Neuromuscular Disorders |
|
| James Grogan,Zachary Simmons | | Current Treatment Options in Neurology. 2020; 22(6) | | [Pubmed] | [DOI] | | | 7 |
Smart wearable system for safety-related medical IoT application: case of epileptic patient working in industrial environment |
|
| Ali Hayek,Samer Telawi,Josef Börcsök,Roy Abi Zeid Daou,Nashaat Halabi | | Health and Technology. 2019; | | [Pubmed] | [DOI] | | | 8 |
Mobiles Anfallsmonitoring bei Epilepsiepatienten |
|
| A. Schulze-Bonhage,S. Böttcher,M. Glasstetter,N. Epitashvili,E. Bruno,M. Richardson,K. v. Laerhoven,M. Dümpelmann | | Der Nervenarzt. 2019; | | [Pubmed] | [DOI] | | | 9 |
Seizure detection devices for use in antiseizure medication clinical trials: A systematic review |
|
| Abhinav V. Kurada,Tarun Srinivasan,Sarah Hammond,Adriana Ulate-Campos,Jonathan Bidwell | | Seizure. 2019; 66: 61 | | [Pubmed] | [DOI] | | | 10 |
Nocturnal Epileptic Seizures Detection Using Inertial and Muscular Sensors |
|
| Osman Salem, Khalid Alsubhi, Ahmed Mehaoua, Raouf Boutaba | | IEEE Transactions on Mobile Computing. 2019; 18(12): 2913 | | [Pubmed] | [DOI] | | | 11 |
Machine learning applications in epilepsy |
|
| Bardia Abbasi,Daniel M. Goldenholz | | Epilepsia. 2019; 60(10): 2037 | | [Pubmed] | [DOI] | | | 12 |
Neural stimulation systems for the control of refractory epilepsy: a review |
|
| Matthew D. Bigelow,Abbas Z. Kouzani | | Journal of NeuroEngineering and Rehabilitation. 2019; 16(1) | | [Pubmed] | [DOI] | | | 13 |
Seizure detection: do current devices work? And when can they be useful? |
|
| Xiuhe Zhao,Samden D. Lhatoo | | Current Neurology and Neuroscience Reports. 2018; 18(7) | | [Pubmed] | [DOI] | | | 14 |
Non-EEG seizure detection systems and potential SUDEP prevention: State of the art |
|
| Anouk Van de Vel,Kris Cuppens,Bert Bonroy,Milica Milosevic,Katrien Jansen,Sabine Van Huffel,Bart Vanrumste,Patrick Cras,Lieven Lagae,Berten Ceulemans | | Seizure. 2016; 41: 141 | | [Pubmed] | [DOI] | |
|
 |
|
|
|
Search Pubmed for