Neuroscience informatics最新文献_第7页

Leveraging transparent ontology learning to refine constructs in neuroscience 利用透明的本体学习来完善神经科学的结构

Neuroscience informatics

Pub Date : 2025-06-01 Epub Date: 2025-03-28 DOI: 10.1016/j.neuri.2025.100199

David Moreau, Kristina Wiebels

引用次数: 0

Automated classification of epileptic seizures using modified one-dimensional convolution neural network based on empirical mode decomposition with high accuracy 基于经验模态分解的改进一维卷积神经网络在癫痫发作自动分类中的应用

Neuroscience informatics

Pub Date : 2025-06-01 Epub Date: 2025-02-13 DOI: 10.1016/j.neuri.2025.100188

Ibtihal Hassan Elshekhidris , Magdi B. M. Amien , Ahmed Fragoon

Background and objectives: The method of electroencephalography (EEG) is frequently employed to identify epileptic seizures. Visually inspecting nonlinear EEG waves is a very difficult and time-consuming process. Therefore, to help with patients' long-term assessment and treatment, an effective automatic detection system is required. Traditional methods of machine learning require step of feature extraction by manual which leads to time consuming, for it we modified in one-Dimensional convolution neural network architecture for features extraction and features dimension reduction for makes the classification low computational complexity and more accurate.

Methods: In this study, we did a comparison between three methods for classification: support vector machine, artificial neural network and one-dimensional convolution neural network. We used the stationary wavelet transform with mother function symlet2 for denoising EEG signal and used the empirical mode decomposition for signal decomposition. After that, features extraction step is necessary when used the support vector machine and artificial neural network, but when use the convolution neural network the features are extracted by layers.

Results: The highest value of a classification accuracy was 100%, and a sensitivity 100%, a specificity 100%, and a precision 100%, which appeared five times when using the one-dimensional convolution neural network after empirical mode decomposition method.

Conclusions: The efficiency of the three methods has been compared and evaluated by using four metrics: Accuracy, Sensitivity, specificity, and Precision, and the result showed the one-dimensional convolution neural network is the best method for classification with empirical mode decomposition.

背景和目的：脑电图（EEG）的方法经常被用来识别癫痫发作。视觉检测非线性脑电波是一个非常困难和耗时的过程。因此，为了帮助患者的长期评估和治疗，需要一个有效的自动检测系统。传统的机器学习方法需要手动进行特征提取，耗时长，在一维卷积神经网络架构上进行特征提取和特征降维，使得分类计算复杂度低，准确率高。方法：对支持向量机、人工神经网络和一维卷积神经网络三种分类方法进行比较。采用带母函数symlet2的平稳小波变换对脑电信号进行降噪，并采用经验模态分解对信号进行分解。在此之后，使用支持向量机和人工神经网络时需要进行特征提取步骤，而使用卷积神经网络时则是逐层提取特征。结果：使用经验模态分解方法后的一维卷积神经网络，分类准确率最高为100%，灵敏度最高为100%，特异性最高为100%，精度最高为100%，出现了5次。结论：通过准确度、灵敏度、特异性和精密度4个指标对3种分类方法的效率进行了比较和评价，结果表明一维卷积神经网络是经验模态分解分类的最佳方法。

{"title":"Automated classification of epileptic seizures using modified one-dimensional convolution neural network based on empirical mode decomposition with high accuracy","authors":"Ibtihal Hassan Elshekhidris , Magdi B. M. Amien , Ahmed Fragoon","doi":"10.1016/j.neuri.2025.100188","DOIUrl":"10.1016/j.neuri.2025.100188","url":null,"abstract":"<div><div>Background and objectives: The method of electroencephalography (EEG) is frequently employed to identify epileptic seizures. Visually inspecting nonlinear EEG waves is a very difficult and time-consuming process. Therefore, to help with patients' long-term assessment and treatment, an effective automatic detection system is required. Traditional methods of machine learning require step of feature extraction by manual which leads to time consuming, for it we modified in one-Dimensional convolution neural network architecture for features extraction and features dimension reduction for makes the classification low computational complexity and more accurate.</div><div>Methods: In this study, we did a comparison between three methods for classification: support vector machine, artificial neural network and one-dimensional convolution neural network. We used the stationary wavelet transform with mother function symlet2 for denoising EEG signal and used the empirical mode decomposition for signal decomposition. After that, features extraction step is necessary when used the support vector machine and artificial neural network, but when use the convolution neural network the features are extracted by layers.</div><div>Results: The highest value of a classification accuracy was 100%, and a sensitivity 100%, a specificity 100%, and a precision 100%, which appeared five times when using the one-dimensional convolution neural network after empirical mode decomposition method.</div><div>Conclusions: The efficiency of the three methods has been compared and evaluated by using four metrics: Accuracy, Sensitivity, specificity, and Precision, and the result showed the one-dimensional convolution neural network is the best method for classification with empirical mode decomposition.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 2","pages":"Article 100188"},"PeriodicalIF":0.0,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143430131","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Unlocking transcranial FUS-EEG feature fusion for non-invasive sleep staging in next-gen clinical applications 解锁经颅FUS-EEG特征融合用于无创睡眠分期的下一代临床应用

Neuroscience informatics

Pub Date : 2025-06-01 Epub Date: 2025-05-07 DOI: 10.1016/j.neuri.2025.100209

Suneet Gupta , Praveen Gupta , Bechoo Lal , Aniruddha Deka , Hirakjyoti Sarma , Sheifali Gupta

Accurate and non-invasive sleep staging is essential for evaluating sleep quality and diagnosing neurological and sleep disorders. Addressing the variations in electroencephalogram (EEG) and electrooculogram (EOG) signals across different sleep stages, this study introduces a transcranial focused ultrasound (tFUS) based multimodal feature fusion deep learning model (MFDL) for automated sleep staging. The proposed framework integrates two one-dimensional convolutional neural networks (1D-CNNs) to extract sleep-relevant features from EEG and EOG signals, followed by an adaptive feature fusion module that dynamically assigns weights based on feature significance. By enhancing discriminative features and suppressing irrelevant ones, the model generates a robust multimodal representation of sleep information. Furthermore, a bidirectional long short-term memory (Bi-LSTM) network captures temporal dependencies in sleep stage transitions, improving classification accuracy. The effectiveness of MFDL is validated on the publicly available Sleep-EDF dataset, achieving 94.1% accuracy, 88.2% Kappa coefficient, and 81.9% MF1 score. Notably, the recall rates for the challenging N1 and REM sleep stages are significantly enhanced to 64.6% and 93.5%, respectively. These results highlight the potential of MFDL in enhancing tFUS-based neuromodulation by providing precise, data-driven sleep state monitoring, paving the way for advanced non-invasive brain stimulation technologies in next-gen clinical applications.

准确和非侵入性的睡眠分期对于评估睡眠质量和诊断神经和睡眠障碍至关重要。针对不同睡眠阶段脑电图（EEG）和眼电图（EOG）信号的变化，本研究引入了一种基于经颅聚焦超声（tFUS）的多模态特征融合深度学习模型（MFDL），用于自动睡眠分期。该框架集成了两个一维卷积神经网络（1d - cnn），从EEG和EOG信号中提取睡眠相关特征，然后采用自适应特征融合模块，根据特征显著性动态分配权重。通过增强判别特征和抑制不相关特征，该模型生成了睡眠信息的鲁棒多模态表示。此外，双向长短期记忆（Bi-LSTM）网络捕获了睡眠阶段转换的时间依赖性，提高了分类准确性。在公开可用的Sleep-EDF数据集上验证了MFDL的有效性，达到94.1%的准确率，88.2%的Kappa系数和81.9%的MF1得分。值得注意的是，具有挑战性的N1和REM睡眠阶段的回忆率显著提高，分别为64.6%和93.5%。这些结果强调了MFDL通过提供精确的、数据驱动的睡眠状态监测来增强基于tfus的神经调节的潜力，为下一代临床应用的先进非侵入性脑刺激技术铺平了道路。

{"title":"Unlocking transcranial FUS-EEG feature fusion for non-invasive sleep staging in next-gen clinical applications","authors":"Suneet Gupta , Praveen Gupta , Bechoo Lal , Aniruddha Deka , Hirakjyoti Sarma , Sheifali Gupta","doi":"10.1016/j.neuri.2025.100209","DOIUrl":"10.1016/j.neuri.2025.100209","url":null,"abstract":"<div><div>Accurate and non-invasive sleep staging is essential for evaluating sleep quality and diagnosing neurological and sleep disorders. Addressing the variations in electroencephalogram (EEG) and electrooculogram (EOG) signals across different sleep stages, this study introduces a transcranial focused ultrasound (tFUS) based multimodal feature fusion deep learning model (MFDL) for automated sleep staging. The proposed framework integrates two one-dimensional convolutional neural networks (1D-CNNs) to extract sleep-relevant features from EEG and EOG signals, followed by an adaptive feature fusion module that dynamically assigns weights based on feature significance. By enhancing discriminative features and suppressing irrelevant ones, the model generates a robust multimodal representation of sleep information. Furthermore, a bidirectional long short-term memory (Bi-LSTM) network captures temporal dependencies in sleep stage transitions, improving classification accuracy. The effectiveness of MFDL is validated on the publicly available Sleep-EDF dataset, achieving 94.1% accuracy, 88.2% Kappa coefficient, and 81.9% MF1 score. Notably, the recall rates for the challenging N1 and REM sleep stages are significantly enhanced to 64.6% and 93.5%, respectively. These results highlight the potential of MFDL in enhancing tFUS-based neuromodulation by providing precise, data-driven sleep state monitoring, paving the way for advanced non-invasive brain stimulation technologies in next-gen clinical applications.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 2","pages":"Article 100209"},"PeriodicalIF":0.0,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144069541","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A MATLAB-based tool for converting fNIRS time-series data to Homer3-compatible formats 一个基于matlab的工具，用于将fNIRS时间序列数据转换为homer3兼容格式

Neuroscience informatics

Pub Date : 2025-06-01 Epub Date: 2025-05-05 DOI: 10.1016/j.neuri.2025.100205

Chao Wang , Xiaojun Cheng , Shichao Liu

Functional Near-Infrared Spectroscopy (fNIRS) is increasingly used in cognitive neuroscience and clinical research, yet preprocessing raw time-series data remains challenging. We introduce a lightweight MATLAB tool to automate the conversion of fNIRS data into Homer3-compatible “*.nirs” format. Our solution targets non-SNIRF raw data and offers a standardized, user-friendly method to streamline fNIRS data preparation. This Technical Note describes the tool's design, workflow, and potential improvements for future development.

功能近红外光谱（fNIRS）在认知神经科学和临床研究中的应用越来越广泛，但原始时间序列数据的预处理仍然具有挑战性。我们介绍了一个轻量级的MATLAB工具来自动将fNIRS数据转换为homer3兼容的“*”。nirs”格式。我们的解决方案针对非snirf原始数据，并提供标准化，用户友好的方法来简化fNIRS数据准备。本技术说明描述了该工具的设计、工作流程和未来开发的潜在改进。

引用次数: 0

Integrating brain-inspired computation with big-data analytics for advanced diagnostics in neuroradiology 集成脑启发计算与大数据分析，用于神经放射学的高级诊断

Neuroscience informatics

Pub Date : 2025-06-01 Epub Date: 2025-04-28 DOI: 10.1016/j.neuri.2025.100202

Senthil Kumar , J. Ramprasath , V. Kalpana , Manikandan Rajagopal , Maheswaran S , Rupesh Gupta

Introduction

Neuroradiology encounters considerable difficulties owing to imaging data's intricacy and high-dimensional characteristics. Conventional diagnostic techniques often encounter challenges regarding precision and scalability, resulting in delays and possible misinterpretations. This paper presents the Big-Data Analytics-based Diagnostics (BDA-D) framework, a revolutionary method using computational models derived from neural architectures and sophisticated analytics to tackle these difficulties.

Methods

The BDA-D architecture utilizes data mining, pattern recognition, and machine learning to glean useful neuroanatomical characteristics from massive datasets. By simulating human thought processes, this method speeds up clinical decision-making and improves diagnostic accuracy. To evaluate the effectiveness of the framework, it is put to the test in a clinical environment.

Results and Discussion

Diagnostic precision, processing speed, and dependability are all enhanced by experimental validation. By detecting even the most minute neuroanatomical changes, BDA-D allows for more accurate diagnosis than traditional approaches. Based on the results, neuroradiologists may improve their practices by using cutting-edge computational methods to close the gap between data-driven analytics and their practical use in the clinic. BDA-D discovers important patterns from high-dimensional neuroimaging data through biologically inspired neural networks, reaching a remarkable diagnosis accuracy of 97.18%. Its 95.42% increase in processing speed allows rapid study of important disorders such as strokes and neurodegenerative diseases. BDA-D reduces inter-observer variability with a dependable value of 94.96%, increasing clinical confidence in AI-assisted diagnosis.

Conclusion

A revolutionary change in neurodiagnostics, the BDA-D framework improves efficiency and reliability. This method has the potential to completely transform neuroradiology by combining big-data analytics with sophisticated computer models. It will allow for more rapid and precise diagnosis.

神经放射学由于成像数据的复杂性和高维特征而遇到相当大的困难。传统的诊断技术经常遇到精度和可扩展性方面的挑战，导致延迟和可能的误解。本文介绍了基于大数据分析的诊断（BDA-D）框架，这是一种革命性的方法，使用源自神经架构和复杂分析的计算模型来解决这些困难。方法BDA-D架构利用数据挖掘、模式识别和机器学习从海量数据集中收集有用的神经解剖学特征。通过模拟人类的思维过程，该方法加快了临床决策，提高了诊断的准确性。为了评估该框架的有效性，在临床环境中对其进行了测试。结果与讨论经实验验证，该方法提高了诊断精度、处理速度和可靠性。通过检测即使是最微小的神经解剖变化，BDA-D允许比传统方法更准确的诊断。基于结果，神经放射学家可以通过使用尖端的计算方法来缩小数据驱动分析与临床实际应用之间的差距，从而改进他们的实践。BDA-D通过生物学启发的神经网络从高维神经成像数据中发现重要模式，达到了97.18%的显著诊断准确率。它的处理速度提高了95.42%，可以快速研究中风和神经退行性疾病等重要疾病。BDA-D降低了观察者之间的可变性，可靠值为94.96%，增加了人工智能辅助诊断的临床可信度。结论BDA-D框架是神经诊断学的革命性变革，提高了效率和可靠性。通过将大数据分析与复杂的计算机模型相结合，这种方法有可能彻底改变神经放射学。它将允许更快速和准确的诊断。

{"title":"Integrating brain-inspired computation with big-data analytics for advanced diagnostics in neuroradiology","authors":"Senthil Kumar , J. Ramprasath , V. Kalpana , Manikandan Rajagopal , Maheswaran S , Rupesh Gupta","doi":"10.1016/j.neuri.2025.100202","DOIUrl":"10.1016/j.neuri.2025.100202","url":null,"abstract":"<div><h3>Introduction</h3><div>Neuroradiology encounters considerable difficulties owing to imaging data's intricacy and high-dimensional characteristics. Conventional diagnostic techniques often encounter challenges regarding precision and scalability, resulting in delays and possible misinterpretations. This paper presents the Big-Data Analytics-based Diagnostics (BDA-D) framework, a revolutionary method using computational models derived from neural architectures and sophisticated analytics to tackle these difficulties.</div></div><div><h3>Methods</h3><div>The BDA-D architecture utilizes data mining, pattern recognition, and machine learning to glean useful neuroanatomical characteristics from massive datasets. By simulating human thought processes, this method speeds up clinical decision-making and improves diagnostic accuracy. To evaluate the effectiveness of the framework, it is put to the test in a clinical environment.</div></div><div><h3>Results and Discussion</h3><div>Diagnostic precision, processing speed, and dependability are all enhanced by experimental validation. By detecting even the most minute neuroanatomical changes, BDA-D allows for more accurate diagnosis than traditional approaches. Based on the results, neuroradiologists may improve their practices by using cutting-edge computational methods to close the gap between data-driven analytics and their practical use in the clinic. BDA-D discovers important patterns from high-dimensional neuroimaging data through biologically inspired neural networks, reaching a remarkable diagnosis accuracy of 97.18%. Its 95.42% increase in processing speed allows rapid study of important disorders such as strokes and neurodegenerative diseases. BDA-D reduces inter-observer variability with a dependable value of 94.96%, increasing clinical confidence in AI-assisted diagnosis.</div></div><div><h3>Conclusion</h3><div>A revolutionary change in neurodiagnostics, the BDA-D framework improves efficiency and reliability. This method has the potential to completely transform neuroradiology by combining big-data analytics with sophisticated computer models. It will allow for more rapid and precise diagnosis.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 2","pages":"Article 100202"},"PeriodicalIF":0.0,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143942951","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Computational intelligence in neuroinformatics: Technologies and data analytics 神经信息学中的计算智能：技术和数据分析

Neuroscience informatics

Pub Date : 2025-03-01 Epub Date: 2025-01-15 DOI: 10.1016/j.neuri.2025.100187

Anand Deshpande , Vania Vieira Estrela , Anitha Jude , Jude Hemanth

引用次数: 0

Design of a computational intelligence system for detection of multiple sclerosis with visual evoked potentials 用视觉诱发电位检测多发性硬化症的计算智能系统设计

Neuroscience informatics

Pub Date : 2025-03-01 Epub Date: 2024-10-30 DOI: 10.1016/j.neuri.2024.100177

Moussa Mohsenpourian , Amir Abolfazl Suratgar , Heidar Ali Talebi , Mahsa Arzani , Abdorreza Naser Moghadasi , Seyed Matin Malakouti , Mohammad Bagher Menhaj

In this study, a new approach for modification of membership functions of a fuzzy inference system (FIS) is demonstrated, in order to serve as a pattern recognition tool for classification of patients diagnosed with multiple sclerosis (MS) from healthy controls (HC) using their visually evoked potential (VEP) recordings. The new approach utilizes Krill Herd (KH) optimization algorithm to modify parameters associated with membership functions of both inputs and outputs of an initial Sugeno-type FIS, while making sure that the error corresponding to training of the network is minimized.

This novel pattern recognition system is applied for classification of VEP signals in 11 MS patients and 11 HC's. A feature extraction routine was performed on the VEP signals, and later substantial features were selected in an optimized feature subset selection scheme employing Ant Colony Optimization (ACO) and Simulated Annealing (SA) algorithms. This alone provided further information regarding clinical value of many previously unused VEP features as an aide for making the diagnosis. The newly designed computational intelligence system is shown to outperform popular classifiers (e.g., multilayer perceptron, support-vector machine, etc.) and was able to distinguish MS patients from HC's with an overall accuracy of 90%.

本研究展示了一种修改模糊推理系统（FIS）隶属函数的新方法，以作为一种模式识别工具，利用视觉诱发电位（VEP）记录对诊断为多发性硬化症（MS）和健康对照（HC）的患者进行分类。该方法利用Krill Herd （KH）优化算法对初始sugeno型FIS的输入和输出的隶属度函数相关参数进行修改，同时确保网络训练对应的误差最小。将该模式识别系统应用于11例MS和11例HC的VEP信号分类。首先对VEP信号进行特征提取，然后采用蚁群优化（ACO）和模拟退火（SA）算法对特征子集进行优化选择。仅这一点就提供了关于许多以前未使用的VEP特征作为辅助诊断的临床价值的进一步信息。新设计的计算智能系统被证明优于流行的分类器（例如，多层感知器，支持向量机等），并且能够以90%的总体准确率区分MS患者和HC患者。

{"title":"Design of a computational intelligence system for detection of multiple sclerosis with visual evoked potentials","authors":"Moussa Mohsenpourian , Amir Abolfazl Suratgar , Heidar Ali Talebi , Mahsa Arzani , Abdorreza Naser Moghadasi , Seyed Matin Malakouti , Mohammad Bagher Menhaj","doi":"10.1016/j.neuri.2024.100177","DOIUrl":"10.1016/j.neuri.2024.100177","url":null,"abstract":"<div><div>In this study, a new approach for modification of membership functions of a fuzzy inference system (FIS) is demonstrated, in order to serve as a pattern recognition tool for classification of patients diagnosed with multiple sclerosis (MS) from healthy controls (HC) using their visually evoked potential (VEP) recordings. The new approach utilizes Krill Herd (KH) optimization algorithm to modify parameters associated with membership functions of both inputs and outputs of an initial Sugeno-type FIS, while making sure that the error corresponding to training of the network is minimized.</div><div>This novel pattern recognition system is applied for classification of VEP signals in 11 MS patients and 11 HC's. A feature extraction routine was performed on the VEP signals, and later substantial features were selected in an optimized feature subset selection scheme employing Ant Colony Optimization (ACO) and Simulated Annealing (SA) algorithms. This alone provided further information regarding clinical value of many previously unused VEP features as an aide for making the diagnosis. The newly designed computational intelligence system is shown to outperform popular classifiers (e.g., multilayer perceptron, support-vector machine, etc.) and was able to distinguish MS patients from HC's with an overall accuracy of 90%.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 1","pages":"Article 100177"},"PeriodicalIF":0.0,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143160180","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

KL-FedDis: A federated learning approach with distribution information sharing using Kullback-Leibler divergence for non-IID data KL-FedDis：一种利用非iid数据的Kullback-Leibler散度实现分布信息共享的联邦学习方法

Neuroscience informatics

Pub Date : 2025-03-01 Epub Date: 2024-11-28 DOI: 10.1016/j.neuri.2024.100182

Md. Rahad , Ruhan Shabab , Mohd. Sultan Ahammad , Md. Mahfuz Reza , Amit Karmaker , Md. Abir Hossain

Data Heterogeneity or Non-IID (non-independent and identically distributed) data identification is one of the prominent challenges in Federated Learning (FL). In Non-IID data, clients have their own local data, which may not be independently and identically distributed. This arises because clients involved in federated learning typically have their own unique, local datasets that vary significantly due to factors like geographical location, user behaviors, or specific contexts. Model divergence is another critical challenge where the local models trained on different clients, data may diverge significantly but making it difficult for the global model to converge. To identify the non-IID data, few federated learning models have been introduced as FedDis, FedProx and FedAvg, but their accuracy is too low. To address the clients Non-IID data along with ensuring privacy, federated learning emerged with appropriate distribution mechanism is an effective solution. In this paper, a modified FedDis learning method called KL-FedDis is proposed, which incorporates Kullback-Leibler (KL) divergence as the regularization technique. KL-FedDis improves accuracy and computation time over the FedDis and FedAvg technique by successfully maintaining the distribution information and encouraging improved collaboration among the local models by utilizing KL divergence.

数据异构或非iid（非独立和同分布）数据识别是联邦学习（FL）中的突出挑战之一。在非iid数据中，客户端有自己的本地数据，这些本地数据可能不是独立的、相同的分布。这是因为参与联邦学习的客户端通常有自己独特的本地数据集，这些数据集由于地理位置、用户行为或特定上下文等因素而变化很大。模型分歧是另一个关键挑战，在不同客户端上训练的局部模型，数据可能会显著分歧，但使全局模型难以收敛。为了识别非iid数据，已经引入了一些联邦学习模型，如FedDis、FedProx和fedag，但它们的准确率太低。为了在解决客户端非iid数据的同时确保隐私，采用适当的分发机制的联邦学习是一种有效的解决方案。本文提出了一种改进的fedis学习方法KL- fedis，该方法将Kullback-Leibler （KL）散度作为正则化技术。KL-FedDis通过利用KL散度成功地维护分布信息和促进局部模型之间的协作，提高了FedDis和fedag技术的准确性和计算时间。

{"title":"KL-FedDis: A federated learning approach with distribution information sharing using Kullback-Leibler divergence for non-IID data","authors":"Md. Rahad , Ruhan Shabab , Mohd. Sultan Ahammad , Md. Mahfuz Reza , Amit Karmaker , Md. Abir Hossain","doi":"10.1016/j.neuri.2024.100182","DOIUrl":"10.1016/j.neuri.2024.100182","url":null,"abstract":"<div><div>Data Heterogeneity or Non-IID (non-independent and identically distributed) data identification is one of the prominent challenges in Federated Learning (FL). In Non-IID data, clients have their own local data, which may not be independently and identically distributed. This arises because clients involved in federated learning typically have their own unique, local datasets that vary significantly due to factors like geographical location, user behaviors, or specific contexts. Model divergence is another critical challenge where the local models trained on different clients, data may diverge significantly but making it difficult for the global model to converge. To identify the non-IID data, few federated learning models have been introduced as FedDis, FedProx and FedAvg, but their accuracy is too low. To address the clients Non-IID data along with ensuring privacy, federated learning emerged with appropriate distribution mechanism is an effective solution. In this paper, a modified FedDis learning method called KL-FedDis is proposed, which incorporates Kullback-Leibler (KL) divergence as the regularization technique. KL-FedDis improves accuracy and computation time over the FedDis and FedAvg technique by successfully maintaining the distribution information and encouraging improved collaboration among the local models by utilizing KL divergence.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 1","pages":"Article 100182"},"PeriodicalIF":0.0,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143160178","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

EEG signal based brain stimulation model to detect epileptic neurological disorders 基于脑电图信号的脑刺激模型检测癫痫性神经系统疾病

Neuroscience informatics

Pub Date : 2025-03-01 Epub Date: 2025-01-14 DOI: 10.1016/j.neuri.2025.100186

Haewon Byeon , Udit Mahajan , Ashish Kumar , V. Rama Krishna , Mukesh Soni , Monika Bansal

Background: Manual visual inspection and analysis of electroencephalogram (EEG) signals of patients are susceptible to the subjective influence of doctors. The introduction of GA-PSO improved the categorization accuracy of both the EP (Evoked potential) and normal groups by automatically screening and optimizing the best feature combination of brain networks. Therefore, selecting effective EEG features for automatic recognition of EP is particularly important for Neuroscience.

New method: A phase synchronization index (PSI) brain stimulation is constructed from multi-channel EEG signals, extracting 15 topological features from the perspectives of network nodes and structural functions. In order to optimize and screen feature combinations in both single and cross-frequency bands, the GA-PSO algorithm is utilized as a feature selection tool for choosing epileptic EEG network features.

Result: Feature combinations are made both within and between bands, and the optimal feature mix is found using the PSO and GA-PSO algorithms. The study found that the GA-PSO algorithm outperformed the PSO algorithm, achieving a higher EP recognition accuracy of 0.9335 under cross-frequency band conditions.

Comparison with existing methods: The results indicate that the introduction of the genetic algorithm enables the GA-PSO algorithm to maintain population diversity and avoid premature convergence to local optima, thereby enhancing the search capabilities of the population.

Conclusion: Based on the findings, topological aspects provide a good way to describe the anomalies in the brain networks of epileptic patients and enhance the classification accuracy through combination, which provides help for pathological research and clinical diagnosis of epilepsy.

背景：人工目视检查和分析患者脑电图信号容易受到医生的主观影响。GA-PSO的引入通过自动筛选和优化脑网络的最佳特征组合，提高了EP（诱发电位）和正常组的分类准确率。因此，选择有效的脑电特征进行脑电图的自动识别对于神经科学来说尤为重要。新方法：利用多通道脑电信号构建一个相同步指数（PSI）脑刺激，从网络节点和结构功能的角度提取15个拓扑特征。为了优化和筛选单频段和跨频段的特征组合，利用GA-PSO算法作为特征选择工具，选择癫痫脑电图网络特征。结果：在频带内和频带间进行了特征组合，利用粒子群算法和ga -粒子群算法找到了最优的特征组合。研究发现GA-PSO算法优于PSO算法，在交叉频带条件下EP识别准确率达到0.9335。与现有方法的比较：结果表明，遗传算法的引入使GA-PSO算法保持种群多样性，避免过早收敛到局部最优，从而增强了种群的搜索能力。结论：基于本研究结果，拓扑学方面为描述癫痫患者脑网络的异常提供了很好的方法，并通过组合提高了分类准确率，为癫痫的病理研究和临床诊断提供了帮助。

{"title":"EEG signal based brain stimulation model to detect epileptic neurological disorders","authors":"Haewon Byeon , Udit Mahajan , Ashish Kumar , V. Rama Krishna , Mukesh Soni , Monika Bansal","doi":"10.1016/j.neuri.2025.100186","DOIUrl":"10.1016/j.neuri.2025.100186","url":null,"abstract":"<div><div><strong>Background:</strong> Manual visual inspection and analysis of electroencephalogram (EEG) signals of patients are susceptible to the subjective influence of doctors. The introduction of GA-PSO improved the categorization accuracy of both the EP (Evoked potential) and normal groups by automatically screening and optimizing the best feature combination of brain networks. Therefore, selecting effective EEG features for automatic recognition of EP is particularly important for Neuroscience.</div><div><strong>New method:</strong> A phase synchronization index (PSI) brain stimulation is constructed from multi-channel EEG signals, extracting 15 topological features from the perspectives of network nodes and structural functions. In order to optimize and screen feature combinations in both single and cross-frequency bands, the GA-PSO algorithm is utilized as a feature selection tool for choosing epileptic EEG network features.</div><div><strong>Result:</strong> Feature combinations are made both within and between bands, and the optimal feature mix is found using the PSO and GA-PSO algorithms. The study found that the GA-PSO algorithm outperformed the PSO algorithm, achieving a higher EP recognition accuracy of 0.9335 under cross-frequency band conditions.</div><div><strong>Comparison with existing methods:</strong> The results indicate that the introduction of the genetic algorithm enables the GA-PSO algorithm to maintain population diversity and avoid premature convergence to local optima, thereby enhancing the search capabilities of the population.</div><div><strong>Conclusion:</strong> Based on the findings, topological aspects provide a good way to describe the anomalies in the brain networks of epileptic patients and enhance the classification accuracy through combination, which provides help for pathological research and clinical diagnosis of epilepsy.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 1","pages":"Article 100186"},"PeriodicalIF":0.0,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143160480","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Training size predictably improves machine learning-based epileptic seizure forecasting from wearables 训练大小可预见地提高了基于机器学习的可穿戴设备癫痫发作预测

Neuroscience informatics

Pub Date : 2025-03-01 Epub Date: 2024-12-09 DOI: 10.1016/j.neuri.2024.100184

Mustafa Halimeh , Michele Jackson , Tobias Loddenkemper , Christian Meisel

Objective: Wrist-worn wearable devices that monitor autonomous nervous system function and movement have shown promise in providing non-invasive, broadly applicable seizure forecasts that increase in accuracy with larger training size. Nevertheless, challenges related to missing validation, small number of enrolled patients, insufficient training data, and lack of patient seizure cycles data hinder its clinical implementation. Here we sought to prospectively validate a previously implemented seizure forecasting algorithm using a larger cohort of pediatric patients with epilepsy (pwe), improve it by including information on seizure cycles, and (3) assess the utility of precise power-laws to predict performance as a function of dataset size.

Methods: We used video-EEG recordings from 166 pwe as ground-truth for seizures, recorded electrodermal activity (EDA), peripheral body temperature (TEMP), blood volume pulse (BVP), accelerometery (ACC) and applied a deep neural LSTM network model (NN) on these data along with information on 24-hour cycles to forecast seizures in a leave-one-subject-out cross validation. Evaluations were made using improvement over chance (IoC) and the Brier skill score (BSS), which measured the improvement of the NN Brier score compared to the Brier score of a rate-matched random (RMR) forecast.

Results: Performance quantified by IoC and BSS increased with training data following precise power-law scaling laws, thereby exceeding prior reported performance levels from smaller datasets. Including information on 24-hour seizure cycles further improved performance. For the largest training set we achieved significant IoC in 68% of pwe, an IoC of 27.3% and a BSS of 0.087.

Interpretation: Our results validate a previous forecast approach and indicate that performance improves predictably as a function of dataset size following power-law scaling.

目的：监测自主神经系统功能和运动的腕戴式可穿戴设备有望提供无创、广泛适用的癫痫发作预测，并随着训练规模的扩大而提高准确性。然而，缺少验证、入选患者数量少、培训数据不足以及缺乏患者癫痫发作周期数据等挑战阻碍了其临床实施。在这里，我们试图使用更大的儿科癫痫患者队列来前瞻性地验证先前实现的癫痫发作预测算法，通过包含癫痫发作周期信息来改进它，并且(3)评估精确幂律的效用，以预测数据集大小的性能。方法：我们使用166 pwe的视频脑电图记录作为癫痫发作的基础事实，记录皮肤电活动（EDA）、外周体温（TEMP）、血容量脉搏（BVP）、加速度计（ACC），并在这些数据上应用深度神经LSTM网络模型（NN）以及24小时周期信息，在留一被试交叉验证中预测癫痫发作。使用改进概率（IoC）和Brier技能评分（BSS）进行评估，BSS衡量了与率匹配随机（RMR）预测的Brier评分相比，NN Brier评分的改善。结果：IoC和BSS量化的性能随着训练数据遵循精确的幂律缩放规律而增加，从而超过先前报道的较小数据集的性能水平。包括24小时癫痫发作周期的信息进一步提高了性能。对于最大的训练集，我们在68%的pwe中实现了显著的IoC， IoC为27.3%，BSS为0.087。解释：我们的结果验证了之前的预测方法，并表明性能作为幂律缩放后数据集大小的函数可预测地提高。

{"title":"Training size predictably improves machine learning-based epileptic seizure forecasting from wearables","authors":"Mustafa Halimeh , Michele Jackson , Tobias Loddenkemper , Christian Meisel","doi":"10.1016/j.neuri.2024.100184","DOIUrl":"10.1016/j.neuri.2024.100184","url":null,"abstract":"<div><div>Objective: Wrist-worn wearable devices that monitor autonomous nervous system function and movement have shown promise in providing non-invasive, broadly applicable seizure forecasts that increase in accuracy with larger training size. Nevertheless, challenges related to missing validation, small number of enrolled patients, insufficient training data, and lack of patient seizure cycles data hinder its clinical implementation. Here we sought to prospectively validate a previously implemented seizure forecasting algorithm using a larger cohort of pediatric patients with epilepsy (pwe), improve it by including information on seizure cycles, and (3) assess the utility of precise power-laws to predict performance as a function of dataset size.</div><div>Methods: We used video-EEG recordings from 166 pwe as ground-truth for seizures, recorded electrodermal activity (EDA), peripheral body temperature (TEMP), blood volume pulse (BVP), accelerometery (ACC) and applied a deep neural LSTM network model (NN) on these data along with information on 24-hour cycles to forecast seizures in a leave-one-subject-out cross validation. Evaluations were made using improvement over chance (IoC) and the Brier skill score (BSS), which measured the improvement of the NN Brier score compared to the Brier score of a rate-matched random (RMR) forecast.</div><div>Results: Performance quantified by IoC and BSS increased with training data following precise power-law scaling laws, thereby exceeding prior reported performance levels from smaller datasets. Including information on 24-hour seizure cycles further improved performance. For the largest training set we achieved significant IoC in 68% of pwe, an IoC of 27.3% and a BSS of 0.087.</div><div>Interpretation: Our results validate a previous forecast approach and indicate that performance improves predictably as a function of dataset size following power-law scaling.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 1","pages":"Article 100184"},"PeriodicalIF":0.0,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143160481","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0