M2SKD : Multi-to-Single Knowledge Distillation of Real-Time Epileptic Seizure Detection for Low-Power Wearable Systems
Baghersalimi, Saleh ; Amirshahi, Alireza ; Forooghifar, Farnaz ; Teijeiro, Tomas ; Aminifar, Amir LU
and Atienza, David
(2024)
In ACM Transactions on Intelligent Systems and Technology
15(5).
- Abstract
Integrating low-power wearable systems into routine health monitoring is an ongoing challenge. Recent advances in the computation capabilities of wearables make it possible to target complex scenarios by exploiting multiple biosignals and using high-performance algorithms, such as Deep Neural Networks (DNNs). However, there is a tradeoff between the algorithms' performance and the low-power requirements of platforms with limited resources. Besides, physically larger and multi-biosignal-based wearables bring significant discomfort to the patients. Consequently, reducing power consumption and discomfort is necessary for patients to use wearable devices continuously during everyday life. To overcome these challenges, in the context of... (More)
Integrating low-power wearable systems into routine health monitoring is an ongoing challenge. Recent advances in the computation capabilities of wearables make it possible to target complex scenarios by exploiting multiple biosignals and using high-performance algorithms, such as Deep Neural Networks (DNNs). However, there is a tradeoff between the algorithms' performance and the low-power requirements of platforms with limited resources. Besides, physically larger and multi-biosignal-based wearables bring significant discomfort to the patients. Consequently, reducing power consumption and discomfort is necessary for patients to use wearable devices continuously during everyday life. To overcome these challenges, in the context of epileptic seizure detection, we propose the Multi-to-Single Knowledge Distillation (M2SKD) approach targeting single-biosignal processing in wearable systems. The starting point is to train a highly-accurate multi-biosignal DNN, then apply M2SKD to develop a single-biosignal DNN solution for wearable systems that achieves an accuracy comparable to the original multi-biosignal DNN. To assess the practicality of our approach to real-life scenarios, we perform a comprehensive simulation experiment analysis on several edge computing platforms.
(Less)
- author
- Baghersalimi, Saleh
; Amirshahi, Alireza
; Forooghifar, Farnaz
; Teijeiro, Tomas
; Aminifar, Amir
LU
and Atienza, David
- organization
- publishing date
- 2024-10-17
- type
- Contribution to journal
- publication status
- published
- subject
- keywords
- deep learning, Edge computing, electrocardiography, epilepsy, knowledge distillation, multi-modal biosignal processing, seizure detection
- in
- ACM Transactions on Intelligent Systems and Technology
- volume
- 15
- issue
- 5
- article number
- 102
- publisher
- Association for Computing Machinery
- external identifiers
-
- scopus:85209888009
- ISSN
- 2157-6904
- DOI
- 10.1145/3675402
- language
- English
- LU publication?
- yes
- additional info
- Publisher Copyright: © 2024 Copyright held by the owner/author(s). Publication rights licensed to ACM.
- id
- a461419a-4a50-4777-9ab5-0eb257518502
- date added to LUP
- 2025-01-15 09:27:01
- date last changed
- 2025-06-04 20:44:18
@article{a461419a-4a50-4777-9ab5-0eb257518502,
abstract = {{<p>Integrating low-power wearable systems into routine health monitoring is an ongoing challenge. Recent advances in the computation capabilities of wearables make it possible to target complex scenarios by exploiting multiple biosignals and using high-performance algorithms, such as Deep Neural Networks (DNNs). However, there is a tradeoff between the algorithms' performance and the low-power requirements of platforms with limited resources. Besides, physically larger and multi-biosignal-based wearables bring significant discomfort to the patients. Consequently, reducing power consumption and discomfort is necessary for patients to use wearable devices continuously during everyday life. To overcome these challenges, in the context of epileptic seizure detection, we propose the Multi-to-Single Knowledge Distillation (M2SKD) approach targeting single-biosignal processing in wearable systems. The starting point is to train a highly-accurate multi-biosignal DNN, then apply M2SKD to develop a single-biosignal DNN solution for wearable systems that achieves an accuracy comparable to the original multi-biosignal DNN. To assess the practicality of our approach to real-life scenarios, we perform a comprehensive simulation experiment analysis on several edge computing platforms.</p>}},
author = {{Baghersalimi, Saleh and Amirshahi, Alireza and Forooghifar, Farnaz and Teijeiro, Tomas and Aminifar, Amir and Atienza, David}},
issn = {{2157-6904}},
keywords = {{deep learning; Edge computing; electrocardiography; epilepsy; knowledge distillation; multi-modal biosignal processing; seizure detection}},
language = {{eng}},
month = {{10}},
number = {{5}},
publisher = {{Association for Computing Machinery}},
series = {{ACM Transactions on Intelligent Systems and Technology}},
title = {{M2SKD : Multi-to-Single Knowledge Distillation of Real-Time Epileptic Seizure Detection for Low-Power Wearable Systems}},
url = {{http://dx.doi.org/10.1145/3675402}},
doi = {{10.1145/3675402}},
volume = {{15}},
year = {{2024}},
}