Article
Engineering, Electrical & Electronic
Sohaib Majzoub, Ahmed Fahmy, Fadi Sibai, Maha Diab, Soliman Mahmoud
Summary: This work investigates epilepsy seizure detection using machine learning. The impact of training and test dataset selection on accuracy and efficacy of CNN prediction is studied. A framework utilizing multiple channels of EEG signals and feature extraction technique is proposed to minimize information loss. Results show that the proposed framework achieves an overall accuracy of 94.44% when the training set contains samples from each patient, and 92.98% when the training set contains a subset of the patient pool.
CIRCUITS SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Chemistry, Multidisciplinary
Marcin Kolodziej, Andrzej Majkowski, Andrzej Rysz
Summary: This article explores the possibilities, issues, and challenges associated with utilizing artificial intelligence for seizure detection using the publicly available iEEG database. It presents standard approaches for analyzing iEEG signals and discusses modern deep learning algorithms. The study shows that CNN and LSTM networks yield significantly better results, and the gradient-weighted class activation mapping algorithm can identify important iEEG signal fragments for seizure detection.
APPLIED SCIENCES-BASEL
(2023)
Article
Clinical Neurology
Una Pale, Tomas Teijeiro, David Atienza
Summary: Long-term monitoring of patients with epilepsy is a challenging problem in engineering. This study proposes a novel semi-supervised learning approach based on multi-centroid HD computing, which shows significantly improved performance in epilepsy detection, especially in cases of data imbalance.
FRONTIERS IN NEUROLOGY
(2022)
Article
Engineering, Biomedical
Xuyang Zhao, Noboru Yoshida, Tetsuya Ueda, Hidenori Sugano, Toshihisa Tanaka
Summary: This study applies commonly used models such as LeNet, VGG, ResNet, and ViT to the EEG image classification task, and solves the problems of data imbalance and model interpretation through data augmentation and model explanation methods. The models achieve good performance in seizure detection and provide visual and quantitative information for clinical experts in diagnosis.
JOURNAL OF NEURAL ENGINEERING
(2023)
Article
Clinical Neurology
Shuang Yu, Rima El Atrache, Jianbin Tang, Michele Jackson, Adam Makarucha, Sarah Cantley, Theodore Sheehan, Solveig Vieluf, Bo Zhang, Jeffrey L. Rogers, Iven Mareels, Stefan Harrer, Tobias Loddenkemper
Summary: This study demonstrates the detection of various seizure types through wearable devices worn on the wrist or ankle, using custom-developed deep-learning models.
Article
Chemistry, Multidisciplinary
Zakareya Lasefr, Khaled Elleithy, Ramasani Rakesh Reddy, Eman Abdelfattah, Miad Faezipour
Summary: This paper studied epileptic seizure detection methods based on EEG signals and proposed an enhanced technique with a mobile application for monitoring the classification of EEG signals. The proposed method achieved high accuracy and outperformed previous studies. It will have significant impacts in the medical field and Human-Computer Interaction fields.
APPLIED SCIENCES-BASEL
(2023)
Article
Engineering, Biomedical
Mona Nasseri, Tal Pal Attia, Boney Joseph, Nicholas M. Gregg, Ewan S. Nurse, Pedro F. Viana, Andreas Schulze-Bonhage, Matthias Duempelmann, Gregory Worrell, Dean R. Freestone, Mark P. Richardson, Benjamin H. Brinkmann
Summary: The study developed an adaptively trained deep neural network algorithm that combined wrist-worn device data and transfer learning method to effectively detect motor and non-motor seizures in various environments. The performance of the algorithm was evaluated multiple times, showing better results in detecting motor seizures.
JOURNAL OF NEURAL ENGINEERING
(2021)
Article
Computer Science, Interdisciplinary Applications
Salim Rukhsar, Anil Kumar Tiwari
Summary: This study introduces a novel lightweight Convolution Transformer model that can effectively detect seizures in cross-patient learning, enhancing performance through the inclusion of inductive biases and attention-based pooling.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2023)
Article
Computer Science, Artificial Intelligence
Qigang Liu, Deming Wang, Yuhang Jia, Suyuan Luo, Chongren Wang
Summary: With the frequent occurrence of cyber-security incidents, intrusion detection system (IDS) has been given more attention, but accurately detecting attacks from traffic data stream is challenging due to the diverse nature of network intrusions and serious imbalanced class distribution. Traditional methods suffer from drawbacks, while the proposed models for intrusion detection outperform state-of-art methods on both binary and multi-class classification, by addressing the issues of feature extraction difficulties and imbalanced data distribution through customized loss function.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Mathematics
Weiyu Zhong, Qiaofeng Wu, Guojun Lu, Yun Xue, Xiaohui Hu
Summary: The study emphasizes the need for more sentiment features from various sources to improve hate speech detection performance. It introduces a keyword-enhanced multi-experts framework highlighting both the key information of the sentence and external sentiment information.
Article
Clinical Neurology
Jonas Munch Nielsen, Ivan C. Zibrandtsen, Paolo Masulli, Torben Lykke Sorensen, Tobias S. Andersen, Troels Wesenberg Kjaer
Summary: This study explores the potential of wearable multi-modal monitoring in epilepsy and identifies effective seizure detection strategies. Automatic seizure detection using multi-modal monitoring shows improved sensitivity and reduced false alarm rates. Visual analysis of multi-modal time series data generates insights for future research on seizure detection.
CLINICAL NEUROPHYSIOLOGY
(2022)
Article
Computer Science, Information Systems
Huaiwen Zhang, Shengsheng Qian, Quan Fang, Changsheng Xu
Summary: With the rapid development of social media platforms and increasing scale of data, rumor detection has become crucial. Existing approaches have limitations in utilizing multi-modal information, feature space difference, and semantic information. This paper proposes a Multi-modal Meta Multi-Task Learning framework, which considers multiple modalities, shares higher meta network-layers, and accurately estimates the weight of each reply using attention mechanism, achieving state-of-the-art performance.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)
Article
Computer Science, Information Systems
Yonghua Yang, Rani A. Sarkis, Rima El Atrache, Tobias Loddenkemper, Christian Meisel
Summary: Automated detection of generalized tonic-clonic seizures (GTCSs) from videos using deep learning has been proven feasible and effective, showing better performance and potential compared to traditional methods. Results demonstrate that deep learning networks based on video sequences outperform detection based on individual frames, with an average sensitivity of 88% and specificity of 92%, and a detection latency of approximately 22 seconds.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2021)
Article
Computer Science, Artificial Intelligence
Loic Jezequel, Ngoc-Son Vu, Jean Beaudet, Aymeric Histace
Summary: In this work, three novel and efficient discriminative and generative tasks are introduced to address the limitations of existing methods in capturing finer features and dealing with fine-grained problems. A new out-of-distribution detection function is proposed and shown to have better stability compared to existing methods. Experimental results demonstrate that our method significantly outperforms state-of-the-art on various anomaly types.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Valentin Gabeff, Tomas Teijeiro, Marina Zapater, Leila Cammoun, Sylvain Rheims, Philippe Ryvlin, David Atienza
Summary: This study addressed the lack of interpretability of neural network models in medical decision support by developing a deep learning model from EEG signals for online detection of epileptic seizures and associating model behavior with expert medical knowledge. The focus was on aggregating classification results, identifying frequency patterns, and recognizing signal waveforms. Results showed that the kernel size in the first layer significantly impacts feature interpretability and model sensitivity.
ARTIFICIAL INTELLIGENCE IN MEDICINE
(2021)