{"title":"AttenEpilepsy:基于多头自我注意的二维卷积网络模型","authors":"","doi":"10.1016/j.enganabound.2024.105989","DOIUrl":null,"url":null,"abstract":"<div><div>The existing epilepsy detection models focus more on local information than the true meaning of long-range dependence when capturing time–frequency image features. This results in imprecise feature vector extraction and room for optimization of detection accuracy. AttenEpilepsy is a novel 2D convolutional network model that uses a multi-head self-attention mechanism to classify epileptic seizure periods, inter-seizure periods, and health states of single-channel EEG signals. The AttenEpilepsy model consists of two parts, namely feature extraction and time–frequency context encoding (STCE). A feature extraction method combining multi-path convolution and adaptive hybrid feature recalibration is proposed, in which multi-path convolution with convolution kernels of different sizes is used to extract relevant multi-scale features from time–frequency images. STCE consists of two modules: multi-head self-attention and causal convolution. A modified multi-head self-attention mechanism is used to model the extracted time–frequency features, and causal convolution is used to analyse the frequency information on the time dependencies. A public dataset from the University of Bonn Epilepsy Research Center is used to evaluate the performance of the AttenEpilepsy model. The experimental results show that the AttenEpilepsy model achieved accuracy (AC), sensitivity (SE), specificity (SP), and F1 score (F1) of 99.81%, 99.82%, 99.89%, and 99.83%, respectively. Further testing of the robustness of the model is conducted by introducing various types of noise into the input data. The proposed AttenEpilepsy network model outperforms the state-of-the-art in terms of various evaluation metrics.</div></div>","PeriodicalId":51039,"journal":{"name":"Engineering Analysis with Boundary Elements","volume":null,"pages":null},"PeriodicalIF":4.2000,"publicationDate":"2024-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"AttenEpilepsy: A 2D convolutional network model based on multi-head self-attention\",\"authors\":\"\",\"doi\":\"10.1016/j.enganabound.2024.105989\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><div>The existing epilepsy detection models focus more on local information than the true meaning of long-range dependence when capturing time–frequency image features. This results in imprecise feature vector extraction and room for optimization of detection accuracy. AttenEpilepsy is a novel 2D convolutional network model that uses a multi-head self-attention mechanism to classify epileptic seizure periods, inter-seizure periods, and health states of single-channel EEG signals. The AttenEpilepsy model consists of two parts, namely feature extraction and time–frequency context encoding (STCE). A feature extraction method combining multi-path convolution and adaptive hybrid feature recalibration is proposed, in which multi-path convolution with convolution kernels of different sizes is used to extract relevant multi-scale features from time–frequency images. STCE consists of two modules: multi-head self-attention and causal convolution. A modified multi-head self-attention mechanism is used to model the extracted time–frequency features, and causal convolution is used to analyse the frequency information on the time dependencies. A public dataset from the University of Bonn Epilepsy Research Center is used to evaluate the performance of the AttenEpilepsy model. The experimental results show that the AttenEpilepsy model achieved accuracy (AC), sensitivity (SE), specificity (SP), and F1 score (F1) of 99.81%, 99.82%, 99.89%, and 99.83%, respectively. Further testing of the robustness of the model is conducted by introducing various types of noise into the input data. The proposed AttenEpilepsy network model outperforms the state-of-the-art in terms of various evaluation metrics.</div></div>\",\"PeriodicalId\":51039,\"journal\":{\"name\":\"Engineering Analysis with Boundary Elements\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":4.2000,\"publicationDate\":\"2024-10-17\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Engineering Analysis with Boundary Elements\",\"FirstCategoryId\":\"5\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S0955799724004624\",\"RegionNum\":2,\"RegionCategory\":\"工程技术\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"ENGINEERING, MULTIDISCIPLINARY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Engineering Analysis with Boundary Elements","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0955799724004624","RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENGINEERING, MULTIDISCIPLINARY","Score":null,"Total":0}
AttenEpilepsy: A 2D convolutional network model based on multi-head self-attention
The existing epilepsy detection models focus more on local information than the true meaning of long-range dependence when capturing time–frequency image features. This results in imprecise feature vector extraction and room for optimization of detection accuracy. AttenEpilepsy is a novel 2D convolutional network model that uses a multi-head self-attention mechanism to classify epileptic seizure periods, inter-seizure periods, and health states of single-channel EEG signals. The AttenEpilepsy model consists of two parts, namely feature extraction and time–frequency context encoding (STCE). A feature extraction method combining multi-path convolution and adaptive hybrid feature recalibration is proposed, in which multi-path convolution with convolution kernels of different sizes is used to extract relevant multi-scale features from time–frequency images. STCE consists of two modules: multi-head self-attention and causal convolution. A modified multi-head self-attention mechanism is used to model the extracted time–frequency features, and causal convolution is used to analyse the frequency information on the time dependencies. A public dataset from the University of Bonn Epilepsy Research Center is used to evaluate the performance of the AttenEpilepsy model. The experimental results show that the AttenEpilepsy model achieved accuracy (AC), sensitivity (SE), specificity (SP), and F1 score (F1) of 99.81%, 99.82%, 99.89%, and 99.83%, respectively. Further testing of the robustness of the model is conducted by introducing various types of noise into the input data. The proposed AttenEpilepsy network model outperforms the state-of-the-art in terms of various evaluation metrics.
期刊介绍:
This journal is specifically dedicated to the dissemination of the latest developments of new engineering analysis techniques using boundary elements and other mesh reduction methods.
Boundary element (BEM) and mesh reduction methods (MRM) are very active areas of research with the techniques being applied to solve increasingly complex problems. The journal stresses the importance of these applications as well as their computational aspects, reliability and robustness.
The main criteria for publication will be the originality of the work being reported, its potential usefulness and applications of the methods to new fields.
In addition to regular issues, the journal publishes a series of special issues dealing with specific areas of current research.
The journal has, for many years, provided a channel of communication between academics and industrial researchers working in mesh reduction methods
Fields Covered:
• Boundary Element Methods (BEM)
• Mesh Reduction Methods (MRM)
• Meshless Methods
• Integral Equations
• Applications of BEM/MRM in Engineering
• Numerical Methods related to BEM/MRM
• Computational Techniques
• Combination of Different Methods
• Advanced Formulations.