TY - JOUR
T1 - rSeiz 2.0
T2 - A Low Latency and Energy-Efficient Seizure Detector in the IoMT
AU - Sayeed, Md Abu
AU - Mohanty, Saraju P.
AU - Kougianos, Elias
N1 - Publisher Copyright:
© 2023, This is a U.S. Government work and not under copyright protection in the US; foreign copyright protection may apply.
PY - 2023/9
Y1 - 2023/9
N2 - Epilepsy affects 50 million people globally, which requires a real-time, low-power, and low-latency seizure detector to address the problem. An IoT-enabled real-time, energy-efficient, and fast seizure detector has been presented. The seizure detector consists of four crucial components: 1. Pulse exclusion mechanism (PEM) or neighborhood component analysis (NCA) for the selection of weighted channels, 2. Optimal feature extraction, 3. RBO optimized κ -nearest neighbor classifier for seizure detection, and 4. Internet of medical things (IoMT) for remote medical services. The proposed system used two models for weighted channel selection: one uses PEM and the other uses NCA. Key features have been extracted from the weighted channels in a specified timing interval, and later, seizure detection is performed using RBO optimized κ -NN classifier. Both software and hardware validation were performed to evaluate the proposed approach. When PEM and RBO are used together, the system’s latency is reduced dramatically while retaining optimal accuracy. The experimental results from software simulation and hardware implementation show that the proposed seizure detector outperforms the current state of the art and provides essential contributions to smart healthcare.
AB - Epilepsy affects 50 million people globally, which requires a real-time, low-power, and low-latency seizure detector to address the problem. An IoT-enabled real-time, energy-efficient, and fast seizure detector has been presented. The seizure detector consists of four crucial components: 1. Pulse exclusion mechanism (PEM) or neighborhood component analysis (NCA) for the selection of weighted channels, 2. Optimal feature extraction, 3. RBO optimized κ -nearest neighbor classifier for seizure detection, and 4. Internet of medical things (IoMT) for remote medical services. The proposed system used two models for weighted channel selection: one uses PEM and the other uses NCA. Key features have been extracted from the weighted channels in a specified timing interval, and later, seizure detection is performed using RBO optimized κ -NN classifier. Both software and hardware validation were performed to evaluate the proposed approach. When PEM and RBO are used together, the system’s latency is reduced dramatically while retaining optimal accuracy. The experimental results from software simulation and hardware implementation show that the proposed seizure detector outperforms the current state of the art and provides essential contributions to smart healthcare.
KW - Channel reduction
KW - EEG
KW - Energy efficiency
KW - Internet-of-medical-things (IoMT)
KW - Low-latency
KW - Pulse exclusion mechanism (PEM)
UR - http://www.scopus.com/inward/record.url?scp=85165297173&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85165297173&partnerID=8YFLogxK
U2 - 10.1007/s42979-023-02051-1
DO - 10.1007/s42979-023-02051-1
M3 - Article
AN - SCOPUS:85165297173
SN - 2662-995X
VL - 4
JO - SN Computer Science
JF - SN Computer Science
IS - 5
M1 - 532
ER -