Journal of Neuroscience Methods最新文献_第5页

PIDGN: An explainable multimodal deep learning framework for early prediction of Parkinson's disease PIDGN：用于帕金森病早期预测的可解释的多模态深度学习框架。

IF 2.7 4区医学 Q2 BIOCHEMICAL RESEARCH METHODS

Journal of Neuroscience Methods

Pub Date : 2025-01-18 DOI: 10.1016/j.jneumeth.2025.110363

Wenjia Li , Quanrui Rao , Shuying Dong , Mengyuan Zhu , Zhen Yang , Xianggeng Huang , Guangchen Liu

Background

Parkinson's disease (PD), the second most common neurodegenerative disease in the world, is usually not diagnosed until the later stages of the disease, when patients might have already missed the best treatment period. Therefore, more effective prediction methods based on artificial intelligence (AI) are needed to assist physicians in timely diagnosis.

New methods

An explainable deep learning-based early Parkinson's disease diagnostic model, Parkinson's Integrative Diagnostic Gated Network (PIDGN), was designed by fusing Single Nucleotide Polymorphism (SNP) and brain sMRI data. Firstly, unimodal internal information was extracted using EmsembleTree dimensionality reduction method, Transformer encoder and 3D ResNet. Secondly, gated attention fusion technique was utilized to explore the inter-modal interactions. Finally, the classification results were output through the fully connected layer. SHapley additive interpretation (SHAP) values and Gradient-weighted Class Activation Mapping (Grad-CAM) techniques were used to help explain the importance of SNPs and brain regions for PD.

Results

The results showed that the PIDGN model achieved the best results with the accuracy of 0.858 and AUROC of 0.897. Top 20 SNPs and the brain regions near the midbrain potentially related to PD were identified using two explainable techniques via SHAP values and Grad-CAM respectively.

Comparison with existing methods and conclusion

The PIDGN model trained by fusing genetic and imaging data outperforms 13 other commonly used unimodal or bimodal models. Explainable PIDGN model helps deepen understanding of several SNPs and sMRI key factors that may affect PD. This study provides a potentially effective solution for automated early diagnosis of PD using AI.

背景：帕金森病（PD）是世界上第二大最常见的神经退行性疾病，通常直到疾病的晚期才被诊断出来，此时患者可能已经错过了最佳治疗期。因此，需要更有效的基于人工智能（AI）的预测方法来协助医生及时诊断。新方法：通过融合单核苷酸多态性（SNP）和大脑sMRI数据，设计了一个可解释的基于深度学习的早期帕金森病诊断模型——帕金森综合诊断门控网络（PIDGN）。首先，利用EmsembleTree降维方法、Transformer编码器和3D ResNet提取单模态内部信息；其次，采用门控注意融合技术对多模态交互进行研究。最后，通过全连通层输出分类结果。使用SHapley加性解释（SHAP）值和梯度加权类激活映射（GradCAM）技术来帮助解释snp和大脑区域对PD的重要性。结果：结果表明，PIDGN模型的准确率为0.858，AUROC为0.897，达到最佳效果。分别通过SHAP值和GradCAM两种可解释的技术确定了前20个snp和中脑附近可能与PD相关的大脑区域。与现有方法的比较和结论：融合遗传和成像数据训练的PIDGN模型优于其他13种常用的单峰或双峰模型。可解释的PIDGN模型有助于加深对可能影响PD的几个snp和sMRI关键因素的理解。本研究为利用人工智能进行PD的自动早期诊断提供了一种潜在的有效解决方案。

{"title":"PIDGN: An explainable multimodal deep learning framework for early prediction of Parkinson's disease","authors":"Wenjia Li , Quanrui Rao , Shuying Dong , Mengyuan Zhu , Zhen Yang , Xianggeng Huang , Guangchen Liu","doi":"10.1016/j.jneumeth.2025.110363","DOIUrl":"10.1016/j.jneumeth.2025.110363","url":null,"abstract":"<div><h3>Background</h3><div>Parkinson's disease (PD), the second most common neurodegenerative disease in the world, is usually not diagnosed until the later stages of the disease, when patients might have already missed the best treatment period. Therefore, more effective prediction methods based on artificial intelligence (AI) are needed to assist physicians in timely diagnosis.</div></div><div><h3>New methods</h3><div>An explainable deep learning-based early Parkinson's disease diagnostic model, Parkinson's Integrative Diagnostic Gated Network (PIDGN), was designed by fusing Single Nucleotide Polymorphism (SNP) and brain sMRI data. Firstly, unimodal internal information was extracted using EmsembleTree dimensionality reduction method, Transformer encoder and 3D ResNet. Secondly, gated attention fusion technique was utilized to explore the inter-modal interactions. Finally, the classification results were output through the fully connected layer. SHapley additive interpretation (SHAP) values and Gradient-weighted Class Activation Mapping (Grad-CAM) techniques were used to help explain the importance of SNPs and brain regions for PD.</div></div><div><h3>Results</h3><div>The results showed that the PIDGN model achieved the best results with the accuracy of 0.858 and AUROC of 0.897. Top 20 SNPs and the brain regions near the midbrain potentially related to PD were identified using two explainable techniques via SHAP values and Grad-CAM respectively.</div></div><div><h3>Comparison with existing methods and conclusion</h3><div>The PIDGN model trained by fusing genetic and imaging data outperforms 13 other commonly used unimodal or bimodal models. Explainable PIDGN model helps deepen understanding of several SNPs and sMRI key factors that may affect PD. This study provides a potentially effective solution for automated early diagnosis of PD using AI.</div></div>","PeriodicalId":16415,"journal":{"name":"Journal of Neuroscience Methods","volume":"415 ","pages":"Article 110363"},"PeriodicalIF":2.7,"publicationDate":"2025-01-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143006822","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A wireless optogenetic setup in freely moving mice for evaluation of cortical spreading depolarization in a chronic disease model 无线光遗传学装置在自由运动小鼠中评估慢性疾病模型中皮质扩张性去极化。

IF 2.7 4区医学 Q2 BIOCHEMICAL RESEARCH METHODS

Journal of Neuroscience Methods

Pub Date : 2025-01-17 DOI: 10.1016/j.jneumeth.2025.110364

Annika Köhne , Simeon O.A. Helgers , Bettina Kewitz , Rieke M. Haupt , Viktoria Oppermann , Franziska Meinert , Renan Sánchez-Porras , Maryam Said , Johannes Woitzik , Patrick Dömer

Background

Spreading depolarization (SD) is an electrophysiological phenomenon of massive neuronal depolarization that occurs in a multitude of brain injuries. Clinical studies and experimental data have linked the occurrence of SDs with secondary brain damage. However, there is a translational gap because of methodological limitations between clinical and experimental approaches focusing on short-term effects. Moreover, usage of highly invasive SD triggers has put into question to what extent SDs themselves or the induction method had caused emergence of tissue damage.

New method

To overcome this gap, we here show the successful realization of an experimental approach for long-term SD induction in a wireless setup of minimal invasive optogenetic stimulation in freely behaving mice.

Results

The proposed method allows for reliable SD induction over the course of three weeks. SD characteristics induced with the wireless setup were comparable to SDs elicited by KCl or cable-bound optogenetic systems. Immunohistological analysis of c-Fos expression revealed neuronal depolarization across the stimulated hemisphere, whereas TUNEL staining revealed no stimulation related apoptosis.

Comparison with existing methods

Optogenetic SD induction so far relied on cable- or fiber-bound systems which restrict experimental possibilities. The proposed model relies on wireless stimulation that allows SD induction in the home cage. In contrast to existing systems, the wireless setup also allows cage enrichment and group housing, therefore allowing behavioral analyses.

Conclusion

This experimental setup has excellent potential to investigate the question of possible long-term SD effects in mouse models of different acute pathologies like traumatic brain injury or migraine.

背景：扩散性去极化（SD）是一种在多种脑损伤中发生的大量神经元去极化的电生理现象。临床研究和实验数据已将SDs的发生与继发性脑损伤联系起来。然而，由于注重短期效果的临床和实验方法之间的方法学限制，存在翻译差距。此外，高侵入性SD触发器的使用让人质疑SD本身或诱导方法在多大程度上导致了组织损伤的出现。新方法：为了克服这一缺陷，我们在这里展示了一种成功实现的实验方法，在自由行为的小鼠中，在无线设置中进行微创光遗传刺激的长期SD诱导。结果：所提出的方法可以在三周内实现可靠的SD诱导。无线设置诱导的SD特性与KCl或电缆绑定光遗传系统诱导的SD相当。c-Fos表达的免疫组织学分析显示受刺激的半球神经元去极化，而TUNEL染色未显示刺激相关的细胞凋亡。与现有方法的比较：迄今为止，光致SD诱导依赖于电缆或光纤束缚系统，这限制了实验的可能性。所提出的模型依赖于无线刺激，允许在家庭笼子中进行SD感应。与现有系统相比，无线设置还允许笼子浓缩和群体住房，因此可以进行行为分析。结论：该实验装置具有很好的潜力来研究创伤性脑损伤或偏头痛等不同急性病理小鼠模型可能存在的长期SD效应问题。

{"title":"A wireless optogenetic setup in freely moving mice for evaluation of cortical spreading depolarization in a chronic disease model","authors":"Annika Köhne , Simeon O.A. Helgers , Bettina Kewitz , Rieke M. Haupt , Viktoria Oppermann , Franziska Meinert , Renan Sánchez-Porras , Maryam Said , Johannes Woitzik , Patrick Dömer","doi":"10.1016/j.jneumeth.2025.110364","DOIUrl":"10.1016/j.jneumeth.2025.110364","url":null,"abstract":"<div><h3>Background</h3><div>Spreading depolarization (SD) is an electrophysiological phenomenon of massive neuronal depolarization that occurs in a multitude of brain injuries. Clinical studies and experimental data have linked the occurrence of SDs with secondary brain damage. However, there is a translational gap because of methodological limitations between clinical and experimental approaches focusing on short-term effects. Moreover, usage of highly invasive SD triggers has put into question to what extent SDs themselves or the induction method had caused emergence of tissue damage.</div></div><div><h3>New method</h3><div>To overcome this gap, we here show the successful realization of an experimental approach for long-term SD induction in a wireless setup of minimal invasive optogenetic stimulation in freely behaving mice.</div></div><div><h3>Results</h3><div>The proposed method allows for reliable SD induction over the course of three weeks. SD characteristics induced with the wireless setup were comparable to SDs elicited by KCl or cable-bound optogenetic systems. Immunohistological analysis of c-Fos expression revealed neuronal depolarization across the stimulated hemisphere, whereas TUNEL staining revealed no stimulation related apoptosis.</div></div><div><h3>Comparison with existing methods</h3><div>Optogenetic SD induction so far relied on cable- or fiber-bound systems which restrict experimental possibilities. The proposed model relies on wireless stimulation that allows SD induction in the home cage. In contrast to existing systems, the wireless setup also allows cage enrichment and group housing, therefore allowing behavioral analyses.</div></div><div><h3>Conclusion</h3><div>This experimental setup has excellent potential to investigate the question of possible long-term SD effects in mouse models of different acute pathologies like traumatic brain injury or migraine.</div></div>","PeriodicalId":16415,"journal":{"name":"Journal of Neuroscience Methods","volume":"415 ","pages":"Article 110364"},"PeriodicalIF":2.7,"publicationDate":"2025-01-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143006635","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Establishing In-vivo brain microdialysis for comparing concentrations of a variety of cortical neurotransmitters in the awake rhesus macaque between different cognitive states 建立体内脑微透析，比较清醒猕猴不同认知状态下多种皮质神经递质的浓度。

IF 2.7 4区医学 Q2 BIOCHEMICAL RESEARCH METHODS

Journal of Neuroscience Methods

Pub Date : 2025-01-09 DOI: 10.1016/j.jneumeth.2025.110361

Stella Mayer , Pankhuri Saxena , Max Arwed Crayen , Stefan Treue

Background

Neuronal activity is modulated by behavior and cognitive processes. The combination of several neurotransmitter systems, acting directly or indirectly on specific populations of neurons, underlie such modulations. Most studies with non-human primates (NHPs) fail to capture this complexity, partly due to the lack of adequate methods for reliably and simultaneously measuring a broad spectrum of neurotransmitters while the animal engages in behavioral tasks.

New Method

To address this gap, we introduce a novel implementation of brain microdialysis (MD), employing semi-chronically implanted guides and probes in awake, behaving NHPs facilitated by removable insets within a standard recording chamber over extrastriate visual cortex (here, the visual middle temporal area (MT)). This approach allows flexible access to diverse brain regions, including areas deep within the sulcus.

Results

Reliable concentration measurements of GABA, glutamate, norepinephrine, epinephrine, dopamine, serotonin, and choline were achieved from small sample volumes (<20 µl) using ultra-performance liquid chromatography with electrospray ionization-mass spectrometry (UPLC-ESI-MS). Comparing two behavioral states – ‘active’ and ‘inactive’, we observe subtle concentration variations between the two behavioral states and a greater variability of concentrations in the active state. Additionally, we find positively and negatively correlated concentration changes for neurotransmitter pairs between the behavioral states.

Conclusions

Therefore, this MD setup allows insights into the neurochemical dynamics in awake primates, facilitating comprehensive investigations into the roles and the complex interplay of neurotransmitters in cognitive and behavioral functions.

背景：神经元活动受行为和认知过程的调节。几种神经递质系统的结合，直接或间接作用于特定的神经元群，是这种调节的基础。大多数对非人类灵长类动物（NHPs）的研究未能捕捉到这种复杂性，部分原因是缺乏足够的方法来可靠地同时测量动物从事行为任务时的广谱神经递质。新方法：为了解决这一差距，我们引入了一种新的脑微透析（MD）实施方法，在清醒状态下使用半慢性植入的导尿管和探针，通过可移动的插入物在标准记录室内的层外视觉皮层（这里，视觉颞中区（MT））上促进NHPs的行为。这种方法可以灵活地进入不同的大脑区域，包括脑沟深处的区域。结果：小样本量的GABA，谷氨酸，去甲肾上腺素，肾上腺素，多巴胺，血清素和胆碱的可靠浓度测量(结论：因此，这种MD设置可以深入了解清醒灵长类动物的神经化学动力学，促进对神经递质在认知和行为功能中的作用和复杂相互作用的全面研究。

{"title":"Establishing In-vivo brain microdialysis for comparing concentrations of a variety of cortical neurotransmitters in the awake rhesus macaque between different cognitive states","authors":"Stella Mayer , Pankhuri Saxena , Max Arwed Crayen , Stefan Treue","doi":"10.1016/j.jneumeth.2025.110361","DOIUrl":"10.1016/j.jneumeth.2025.110361","url":null,"abstract":"<div><h3>Background</h3><div>Neuronal activity is modulated by behavior and cognitive processes. The combination of several neurotransmitter systems, acting directly or indirectly on specific populations of neurons, underlie such modulations. Most studies with non-human primates (NHPs) fail to capture this complexity, partly due to the lack of adequate methods for reliably and simultaneously measuring a broad spectrum of neurotransmitters while the animal engages in behavioral tasks.</div></div><div><h3>New Method</h3><div>To address this gap, we introduce a novel implementation of brain microdialysis (MD), employing semi-chronically implanted guides and probes in awake, behaving NHPs facilitated by removable insets within a standard recording chamber over extrastriate visual cortex (here, the visual middle temporal area (MT)). This approach allows flexible access to diverse brain regions, including areas deep within the sulcus.</div></div><div><h3>Results</h3><div>Reliable concentration measurements of GABA, glutamate, norepinephrine, epinephrine, dopamine, serotonin, and choline were achieved from small sample volumes (<20 µl) using ultra-performance liquid chromatography with electrospray ionization-mass spectrometry (UPLC-ESI-MS). Comparing two behavioral states – ‘active’ and ‘inactive’, we observe subtle concentration variations between the two behavioral states and a greater variability of concentrations in the active state. Additionally, we find positively and negatively correlated concentration changes for neurotransmitter pairs between the behavioral states.</div></div><div><h3>Conclusions</h3><div>Therefore, this MD setup allows insights into the neurochemical dynamics in awake primates, facilitating comprehensive investigations into the roles and the complex interplay of neurotransmitters in cognitive and behavioral functions.</div></div>","PeriodicalId":16415,"journal":{"name":"Journal of Neuroscience Methods","volume":"415 ","pages":"Article 110361"},"PeriodicalIF":2.7,"publicationDate":"2025-01-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142971275","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

An EEG-based emotion recognition method by fusing multi-frequency-spatial features under multi-frequency bands 基于脑电图的多频段多频空间特征融合情感识别方法。

IF 2.7 4区医学 Q2 BIOCHEMICAL RESEARCH METHODS

Journal of Neuroscience Methods

Pub Date : 2025-01-06 DOI: 10.1016/j.jneumeth.2025.110360

Qiuyu Chen, Xiaoqian Mao, Yuebin Song, Kefa Wang

Background

Recognition of emotion changes is of great significance to a person’s physical and mental health. At present, EEG-based emotion recognition methods are mainly focused on time or frequency domains, but rarely on spatial information. Therefore, the goal of this study is to improve the performance of emotion recognition by integrating frequency and spatial domain information under multi-frequency bands.

New methods

Firstly, EEG signals of four frequency bands are extracted, and then three frequency-spatial features of differential entropy (DE) symmetric difference (SD) and symmetric quotient (SQ) are separately calculated. Secondly, according to the distribution of EEG electrodes, a series of brain maps are constructed by three frequency-spatial features for each frequency band. Thirdly, a Multi-Parallel-Input Convolutional Neural Network (MPICNN) uses the constructed brain maps to train and obtain the emotion recognition model. Finally, the subject-dependent experiments are conducted on DEAP and SEED-IV datasets.

Results

The experimental results of DEAP dataset show that the average accuracy of four-class emotion recognition, namely, high-valence high-arousal, high-valence low-arousal, low-valence high-arousal and low-valence low-arousal, reaches 98.71 %. The results of SEED-IV dataset show the average accuracy of four-class emotion recognition, namely, happy, sad, neutral and fear reaches 92.55 %.

Comparison with existing methods

This method has a best classification performance compared with the state-of-the-art methods on both four-class emotion recognition datasets.

Conclusions

This EEG-based emotion recognition method fused multi-frequency-spatial features under multi-frequency bands, and effectively improved the recognition performance compared with the existing methods.

背景：情绪变化的识别对一个人的身心健康具有重要意义。目前，基于脑电图的情绪识别方法主要集中在时域或频域，很少涉及空间信息。因此，本研究的目标是通过在多频段下整合频率和空间域信息来提高情绪识别的性能。新方法：首先提取四个频段的脑电信号，然后分别计算微分熵（DE）、对称差（SD）和对称商（SQ）三个频率空间特征；其次，根据脑电电极的分布，利用三个频率空间特征对每个频带构造一系列脑图；第三，利用构建的脑图进行多并行输入卷积神经网络（MPICNN）训练，得到情绪识别模型。最后，在DEAP和SEED-IV数据集上进行受试者相关实验。结果：DEAP数据集的实验结果表明，高价高唤醒、高价低唤醒、低价高唤醒和低价低唤醒四类情绪识别的平均正确率达到98.71%。SEED-IV数据集的结果显示，快乐、悲伤、中性和恐惧四类情绪识别的平均准确率达到92.55%。与现有方法的比较：在四类情绪识别数据集上，与现有方法相比，该方法具有最好的分类性能。结论：基于脑电图的情感识别方法在多频带下融合了多频空间特征，与现有方法相比，有效提高了识别性能。

{"title":"An EEG-based emotion recognition method by fusing multi-frequency-spatial features under multi-frequency bands","authors":"Qiuyu Chen, Xiaoqian Mao, Yuebin Song, Kefa Wang","doi":"10.1016/j.jneumeth.2025.110360","DOIUrl":"10.1016/j.jneumeth.2025.110360","url":null,"abstract":"<div><h3>Background</h3><div>Recognition of emotion changes is of great significance to a person’s physical and mental health. At present, EEG-based emotion recognition methods are mainly focused on time or frequency domains, but rarely on spatial information. Therefore, the goal of this study is to improve the performance of emotion recognition by integrating frequency and spatial domain information under multi-frequency bands.</div></div><div><h3>New methods</h3><div>Firstly, EEG signals of four frequency bands are extracted, and then three frequency-spatial features of differential entropy (DE) symmetric difference (SD) and symmetric quotient (SQ) are separately calculated. Secondly, according to the distribution of EEG electrodes, a series of brain maps are constructed by three frequency-spatial features for each frequency band. Thirdly, a Multi-Parallel-Input Convolutional Neural Network (MPICNN) uses the constructed brain maps to train and obtain the emotion recognition model. Finally, the subject-dependent experiments are conducted on DEAP and SEED-IV datasets.</div></div><div><h3>Results</h3><div>The experimental results of DEAP dataset show that the average accuracy of four-class emotion recognition, namely, high-valence high-arousal, high-valence low-arousal, low-valence high-arousal and low-valence low-arousal, reaches 98.71 %. The results of SEED-IV dataset show the average accuracy of four-class emotion recognition, namely, happy, sad, neutral and fear reaches 92.55 %.</div></div><div><h3>Comparison with existing methods</h3><div>This method has a best classification performance compared with the state-of-the-art methods on both four-class emotion recognition datasets.</div></div><div><h3>Conclusions</h3><div>This EEG-based emotion recognition method fused multi-frequency-spatial features under multi-frequency bands, and effectively improved the recognition performance compared with the existing methods.</div></div>","PeriodicalId":16415,"journal":{"name":"Journal of Neuroscience Methods","volume":"415 ","pages":"Article 110360"},"PeriodicalIF":2.7,"publicationDate":"2025-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142950262","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Convolutional Neural Networks for the segmentation of hippocampal structures in postmortem MRI scans 卷积神经网络对死后MRI扫描海马结构的分割。

IF 2.7 4区医学 Q2 BIOCHEMICAL RESEARCH METHODS

Journal of Neuroscience Methods

Pub Date : 2025-01-02 DOI: 10.1016/j.jneumeth.2024.110359

Anoop B.N. , Karl Li , Nicolas Honnorat , Tanweer Rashid , Di Wang , Jinqi Li , Elyas Fadaee , Sokratis Charisis , Jamie M. Walker , Timothy E. Richardson , David A. Wolk , Peter T. Fox , José E. Cavazos , Sudha Seshadri , Laura E.M. Wisse , Mohamad Habes

Background:

The hippocampus plays a crucial role in memory and is one of the first structures affected by Alzheimer’s disease. Postmortem MRI offers a way to quantify the alterations by measuring the atrophy of the inner structures of the hippocampus. Unfortunately, the manual segmentation of hippocampal subregions required to carry out these measures is very time-consuming.

New Method:

In this study, we explore the use of fully automated methods relying on state-of-the-art Deep Learning approaches to produce these annotations. More specifically, we propose a new segmentation framework made of a set of encoder–decoder blocks embedding self-attention mechanisms and atrous spatial pyramidal pooling to produce better maps of the hippocampus and identify four hippocampal regions: the dentate gyrus, the hippocampal head, the hippocampal body, and the hippocampal tail.

Results:

Trained using slices extracted from 15 postmortem T1-weighted, T2-weighted, and susceptibility-weighted MRI scans, our new approach produces hippocampus parcellations that are better aligned with the manually delineated parcellations provided by neuroradiologists.

Comparison with Existing Methods:

Four standard deep learning segmentation architectures: UNet, Double UNet, Attention UNet, and Multi-resolution UNet have been utilized for the qualitative and quantitative comparison of the proposed hippocampal region segmentation model.

Conclusions:

Postmortem MRI serves as a highly valuable neuroimaging technique for examining the effects of neurodegenerative diseases on the intricate structures within the hippocampus. This study opens the way to large sample-size postmortem studies of the hippocampal substructures.

背景：海马体在记忆中起着至关重要的作用，是最早受阿尔茨海默病影响的结构之一。死后MRI通过测量海马内部结构的萎缩，提供了一种量化改变的方法。不幸的是，手工分割海马亚区需要执行这些措施是非常耗时的。新方法：在本研究中，我们探索了依靠最先进的深度学习方法来生成这些注释的全自动方法的使用。更具体地说，我们提出了一种新的分割框架，该框架由一组嵌入自我注意机制的编码器-解码器块和复杂的空间锥体池组成，以产生更好的海马体地图，并识别出四个海马体区域：齿状回、海马头、海马体和海马尾。结果：使用15个尸检t1加权、t2加权和敏感性加权MRI扫描提取的切片进行训练，我们的新方法产生的海马包裹体与神经放射学家提供的人工描绘的包裹体更好地一致。与现有方法的比较：采用四种标准的深度学习分割架构：UNet、Double UNet、Attention UNet和Multi-resolution UNet，对提出的海马区域分割模型进行定性和定量比较。结论：死后MRI是一种非常有价值的神经成像技术，可用于检查神经退行性疾病对海马内复杂结构的影响。这项研究为海马亚结构的大样本死后研究开辟了道路。

{"title":"Convolutional Neural Networks for the segmentation of hippocampal structures in postmortem MRI scans","authors":"Anoop B.N. , Karl Li , Nicolas Honnorat , Tanweer Rashid , Di Wang , Jinqi Li , Elyas Fadaee , Sokratis Charisis , Jamie M. Walker , Timothy E. Richardson , David A. Wolk , Peter T. Fox , José E. Cavazos , Sudha Seshadri , Laura E.M. Wisse , Mohamad Habes","doi":"10.1016/j.jneumeth.2024.110359","DOIUrl":"10.1016/j.jneumeth.2024.110359","url":null,"abstract":"<div><h3>Background:</h3><div>The hippocampus plays a crucial role in memory and is one of the first structures affected by Alzheimer’s disease. Postmortem MRI offers a way to quantify the alterations by measuring the atrophy of the inner structures of the hippocampus. Unfortunately, the manual segmentation of hippocampal subregions required to carry out these measures is very time-consuming.</div></div><div><h3>New Method:</h3><div>In this study, we explore the use of fully automated methods relying on state-of-the-art Deep Learning approaches to produce these annotations. More specifically, we propose a new segmentation framework made of a set of encoder–decoder blocks embedding self-attention mechanisms and atrous spatial pyramidal pooling to produce better maps of the hippocampus and identify four hippocampal regions: the dentate gyrus, the hippocampal head, the hippocampal body, and the hippocampal tail.</div></div><div><h3>Results:</h3><div>Trained using slices extracted from 15 postmortem T1-weighted, T2-weighted, and susceptibility-weighted MRI scans, our new approach produces hippocampus parcellations that are better aligned with the manually delineated parcellations provided by neuroradiologists.</div></div><div><h3>Comparison with Existing Methods:</h3><div>Four standard deep learning segmentation architectures: UNet, Double UNet, Attention UNet, and Multi-resolution UNet have been utilized for the qualitative and quantitative comparison of the proposed hippocampal region segmentation model.</div></div><div><h3>Conclusions:</h3><div>Postmortem MRI serves as a highly valuable neuroimaging technique for examining the effects of neurodegenerative diseases on the intricate structures within the hippocampus. This study opens the way to large sample-size postmortem studies of the hippocampal substructures.</div></div>","PeriodicalId":16415,"journal":{"name":"Journal of Neuroscience Methods","volume":"415 ","pages":"Article 110359"},"PeriodicalIF":2.7,"publicationDate":"2025-01-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142927256","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

An adaptive protocol to assess physiological responses as a function of task demand in speech-in-noise testing 语音噪声测试中任务需求对生理反应的影响。

IF 2.7 4区医学 Q2 BIOCHEMICAL RESEARCH METHODS

Journal of Neuroscience Methods

Pub Date : 2024-12-31 DOI: 10.1016/j.jneumeth.2024.110348

Edoardo Maria Polo , Davide Simeone , Maximiliano Mollura , Alessia Paglialonga , Riccardo Barbieri

Background:

Acoustic challenges impose demands on cognitive resources, known as listening effort (LE), which can substantially influence speech perception and communication. Standardized assessment protocols for monitoring LE are lacking, hindering the development of adaptive hearing assistive technology.

New Method:

We employed an adaptive protocol, including a speech-in-noise test and personalized definition of task demand, to assess LE and its physiological correlates. Features extracted from electroencephalogram, galvanic skin response, electrocardiogram, respiration, pupil dilation, and blood volume pulse responses were analyzed as a function of task demand in 21 healthy participants with normal hearing.

Results:

Heightened sympathetic response was observed with higher task demand, evidenced by increased heart rate, blood pressure, and breath amplitude. Blood volume amplitude and breath amplitude exhibited higher sensitivity to changes in task demand.

Comparison with Existing Methods:

Notably, galvanic skin response showed higher amplitude during low task demand phases, indicating increased attention and engagement, aligning with findings from electroencephalogram signals and Lacey’s attention theory.

Conclusions:

The analysis of a range of physiological signals, spanning cardiovascular, central, and autonomic domains, demonstrated effectiveness in comprehensively examining LE. Future research should explore additional levels and manipulations of task demand, as well as the influence of individual motivation and hearing sensitivity, to further validate these outcomes and enhance the development of adaptive hearing assistive technology.

背景：声学挑战对认知资源提出了要求，即所谓的聆听努力（LE），这会严重影响言语感知和交流。目前还缺乏监测聆听强度的标准化评估方案，这阻碍了自适应听力辅助技术的发展：新方法：我们采用了一种自适应方案，包括噪声言语测试和任务需求的个性化定义，来评估聆听强度及其生理相关性。我们分析了从 21 名听力正常的健康参与者的脑电图、皮肤电反应、心电图、呼吸、瞳孔放大和血容量脉搏反应中提取的特征，并将其作为任务需求的函数：结果：任务要求越高，交感神经反应越强烈，表现为心率、血压和呼吸幅度增大。血容量振幅和呼吸振幅对任务需求变化的敏感性更高：值得注意的是，皮肤电化反应在低任务需求阶段显示出更高的振幅，表明注意力和参与度增加，这与脑电信号和雷西注意力理论的研究结果一致：对心血管、中枢和自律神经领域的一系列生理信号进行分析，证明了全面检查 LE 的有效性。未来的研究应探索任务需求的更多层次和操作方法，以及个人动机和听力敏感度的影响，以进一步验证这些结果，促进自适应听力辅助技术的发展。

{"title":"An adaptive protocol to assess physiological responses as a function of task demand in speech-in-noise testing","authors":"Edoardo Maria Polo , Davide Simeone , Maximiliano Mollura , Alessia Paglialonga , Riccardo Barbieri","doi":"10.1016/j.jneumeth.2024.110348","DOIUrl":"10.1016/j.jneumeth.2024.110348","url":null,"abstract":"<div><h3>Background:</h3><div>Acoustic challenges impose demands on cognitive resources, known as listening effort (LE), which can substantially influence speech perception and communication. Standardized assessment protocols for monitoring LE are lacking, hindering the development of adaptive hearing assistive technology.</div></div><div><h3>New Method:</h3><div>We employed an adaptive protocol, including a speech-in-noise test and personalized definition of task demand, to assess LE and its physiological correlates. Features extracted from electroencephalogram, galvanic skin response, electrocardiogram, respiration, pupil dilation, and blood volume pulse responses were analyzed as a function of task demand in 21 healthy participants with normal hearing.</div></div><div><h3>Results:</h3><div>Heightened sympathetic response was observed with higher task demand, evidenced by increased heart rate, blood pressure, and breath amplitude. Blood volume amplitude and breath amplitude exhibited higher sensitivity to changes in task demand.</div></div><div><h3>Comparison with Existing Methods:</h3><div>Notably, galvanic skin response showed higher amplitude during low task demand phases, indicating increased attention and engagement, aligning with findings from electroencephalogram signals and Lacey’s attention theory.</div></div><div><h3>Conclusions:</h3><div>The analysis of a range of physiological signals, spanning cardiovascular, central, and autonomic domains, demonstrated effectiveness in comprehensively examining LE. Future research should explore additional levels and manipulations of task demand, as well as the influence of individual motivation and hearing sensitivity, to further validate these outcomes and enhance the development of adaptive hearing assistive technology.</div></div>","PeriodicalId":16415,"journal":{"name":"Journal of Neuroscience Methods","volume":"415 ","pages":"Article 110348"},"PeriodicalIF":2.7,"publicationDate":"2024-12-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142920571","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Emotion recognition based on EEG source signals and dynamic brain function network

IF 2.7 4区医学 Q2 BIOCHEMICAL RESEARCH METHODS

Journal of Neuroscience Methods

Pub Date : 2024-12-28 DOI: 10.1016/j.jneumeth.2024.110358

He Sun , Hailing Wang , Raofen Wang , Yufei Gao

Background

Brain network features contain more emotion-related information and can be more effective in emotion recognition. However, emotions change continuously and dynamically, and current function brain network features using the sliding window method cannot explore dynamic characteristics of different emotions, which leads to the serious loss of functional connectivity information.

New method

In the study, we proposed a new framework based on EEG source signals and dynamic function brain network (dyFBN) for emotion recognition. We constructed emotion-related dyFBN with dynamic phase linearity measurement (dyPLM) at every time point and extracted the second-order feature Root Mean Square (RMS) based on of dyFBN. In addition, a multiple feature fusion strategy was employed, integrating sensor frequency features with connection information.

Results

The recognition accuracy of subject-independent and subject-dependent is 83.50 % and 88.93 %, respectively. The selected optimal feature subset of fused features highlighted the interplay between dynamic features and sensor features and showcased the crucial brain regions of the right superiortemporal, left isthmuscingulate, and left parsorbitalis in emotion recognition.

Comparison with existing methods

Compared with current methods, the emotion recognition accuracy of subject-independent and subject-dependent is improved by 11.46 % and 10.19 %, respectively. In addition, recognition accuracy of the fused features of RMS and sensor features is also better than the fused features of existing methods.

Conclusions

These findings prove the validity of the proposed framework, which leads to better emotion recognition.

{"title":"Emotion recognition based on EEG source signals and dynamic brain function network","authors":"He Sun , Hailing Wang , Raofen Wang , Yufei Gao","doi":"10.1016/j.jneumeth.2024.110358","DOIUrl":"10.1016/j.jneumeth.2024.110358","url":null,"abstract":"<div><h3>Background</h3><div>Brain network features contain more emotion-related information and can be more effective in emotion recognition. However, emotions change continuously and dynamically, and current function brain network features using the sliding window method cannot explore dynamic characteristics of different emotions, which leads to the serious loss of functional connectivity information.</div></div><div><h3>New method</h3><div>In the study, we proposed a new framework based on EEG source signals and dynamic function brain network (dyFBN) for emotion recognition. We constructed emotion-related dyFBN with dynamic phase linearity measurement (dyPLM) at every time point and extracted the second-order feature Root Mean Square (RMS) based on of dyFBN. In addition, a multiple feature fusion strategy was employed, integrating sensor frequency features with connection information.</div></div><div><h3>Results</h3><div>The recognition accuracy of subject-independent and subject-dependent is 83.50 % and 88.93 %, respectively. The selected optimal feature subset of fused features highlighted the interplay between dynamic features and sensor features and showcased the crucial brain regions of the right superiortemporal, left isthmuscingulate, and left parsorbitalis in emotion recognition.</div></div><div><h3>Comparison with existing methods</h3><div>Compared with current methods, the emotion recognition accuracy of subject-independent and subject-dependent is improved by 11.46 % and 10.19 %, respectively. In addition, recognition accuracy of the fused features of RMS and sensor features is also better than the fused features of existing methods.</div></div><div><h3>Conclusions</h3><div>These findings prove the validity of the proposed framework, which leads to better emotion recognition.</div></div>","PeriodicalId":16415,"journal":{"name":"Journal of Neuroscience Methods","volume":"415 ","pages":"Article 110358"},"PeriodicalIF":2.7,"publicationDate":"2024-12-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143177174","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A new spatial contrast coding approach for SSVEP-based BCIs

IF 2.7 4区医学 Q2 BIOCHEMICAL RESEARCH METHODS

Journal of Neuroscience Methods

Pub Date : 2024-12-27 DOI: 10.1016/j.jneumeth.2024.110357

Hui Zhong , Gege Ming , Weihua Pei , Xiaorong Gao , Yijun Wang

Background

Steady-state visual evoked potential (SSVEP)-based brain-computer interface (BCI) systems mainly adopt the frequency, phase, and hybrid coding approaches in previous studies. This study proposes a new encoding approach based on spatial contrast, which is one of the spatial properties of visual stimuli.

New method

First, this study designed checkerboard-like stimuli with 11 kinds of background contrast to explore the effect of background contrast on stimulus-response characteristics of SSVEPs. Based on the spatial contrast related modulation of responses, this study conducted offline simulations to evaluate the feasibility of a multi-target contrast coding approach. Finally, this study designed a four-target SSVEP-BCI system to demonstrate the contrast coding approach.

Results

Checkerboard-like stimuli with the same frequency and initial phase but different background contrasts have different SSVEP responses in terms of amplitude, topography, and phase. Taking advantage of the characteristics, both offline simulations and online verifications indicated that the proposed BCI system achieved good classification performance. Online BCI experiments found that the four-target SSVEP-BCI system achieved averaged information transfer rates of 59.58 ± 0.42 bits/min at the 15 Hz condition and 52.54 ± 2.32 bits/min at the 30 Hz condition, respectively.

Comparison with existing method

Different from previous frequency, phase, and spatial coding approaches, this study adopts a background contrast-based coding approach to achieve a four-target BCI system.

Conclusion

This study proposes a new spatial contrast coding approach, which will enrich the encoding approach of the SSVEP-BCI systems and promote the applications of the SSVEP-BCI systems in more scenarios.

{"title":"A new spatial contrast coding approach for SSVEP-based BCIs","authors":"Hui Zhong , Gege Ming , Weihua Pei , Xiaorong Gao , Yijun Wang","doi":"10.1016/j.jneumeth.2024.110357","DOIUrl":"10.1016/j.jneumeth.2024.110357","url":null,"abstract":"<div><h3>Background</h3><div>Steady-state visual evoked potential (SSVEP)-based brain-computer interface (BCI) systems mainly adopt the frequency, phase, and hybrid coding approaches in previous studies. This study proposes a new encoding approach based on spatial contrast, which is one of the spatial properties of visual stimuli.</div></div><div><h3>New method</h3><div>First, this study designed checkerboard-like stimuli with 11 kinds of background contrast to explore the effect of background contrast on stimulus-response characteristics of SSVEPs. Based on the spatial contrast related modulation of responses, this study conducted offline simulations to evaluate the feasibility of a multi-target contrast coding approach. Finally, this study designed a four-target SSVEP-BCI system to demonstrate the contrast coding approach.</div></div><div><h3>Results</h3><div>Checkerboard-like stimuli with the same frequency and initial phase but different background contrasts have different SSVEP responses in terms of amplitude, topography, and phase. Taking advantage of the characteristics, both offline simulations and online verifications indicated that the proposed BCI system achieved good classification performance. Online BCI experiments found that the four-target SSVEP-BCI system achieved averaged information transfer rates of 59.58 ± 0.42 bits/min at the 15 Hz condition and 52.54 ± 2.32 bits/min at the 30 Hz condition, respectively.</div></div><div><h3>Comparison with existing method</h3><div>Different from previous frequency, phase, and spatial coding approaches, this study adopts a background contrast-based coding approach to achieve a four-target BCI system.</div></div><div><h3>Conclusion</h3><div>This study proposes a new spatial contrast coding approach, which will enrich the encoding approach of the SSVEP-BCI systems and promote the applications of the SSVEP-BCI systems in more scenarios.</div></div>","PeriodicalId":16415,"journal":{"name":"Journal of Neuroscience Methods","volume":"415 ","pages":"Article 110357"},"PeriodicalIF":2.7,"publicationDate":"2024-12-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143177175","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Estimation of fiber orientation distributions in superficial white matter using an asymmetric constrained spherical deconvolution method

IF 2.7 4区医学 Q2 BIOCHEMICAL RESEARCH METHODS

Journal of Neuroscience Methods

Pub Date : 2024-12-25 DOI: 10.1016/j.jneumeth.2024.110353

Jingxin Meng, Jianglin He, Yuanjun Wang

Background

Superficial white matter is an important component of white matter. Estimation of fiber orientation distributions based on diffusion magnetic resonance imaging is a critical step in white matter tractography imaging. However, due to the complex structure of superficial white matter, existing models for estimating fiber orientation distributions are ineffective in reconstructing superficial white matter and even reconstruct incorrect orientation distributions.

New method

In this paper, we improve the traditional constrained spherical deconvolution method and propose a novel asymmetric constrained spherical deconvolution method. The method takes into account that the displacement profile of the water molecules in brain tissue are non-Gaussian diffusion and the core parameter kurtosis might characterize tissue structure better than diffusivity coefficients. So diffusion kurtosis imaging model is used to estimate the white matter response function. The proposed method applies the diffusion kurtosis imaging model response function to the asymmetric fiber orientation distributions, and this is the first attempt to obtain more accurate fiber orientation distributions. Furthermore, the Gaussian-Distribution distance weight and Watson-Distribution angle weight are used for asymmetric regularization.

Results

We evaluate the method using FiberCup phantom, ISMRM 2015 data and in vivo data provided CHCP dataset. The results show that our proposed method can more accurately reconstruct the complex fiber structure of superficial white matter with more accurate fiber orientation, fewer pseudo-peaks, and mitigate gyral bias.

Comparison with existing methods

Our proposed method has higher accuracy in estimating the fiber orientation distributions and can reconstruct highly curved fiber voxels.

Conclusion

This proposed method provides new insights into the estimation of the orientation distribution of superficial white matter fibers.

{"title":"Estimation of fiber orientation distributions in superficial white matter using an asymmetric constrained spherical deconvolution method","authors":"Jingxin Meng, Jianglin He, Yuanjun Wang","doi":"10.1016/j.jneumeth.2024.110353","DOIUrl":"10.1016/j.jneumeth.2024.110353","url":null,"abstract":"<div><h3>Background</h3><div>Superficial white matter is an important component of white matter. Estimation of fiber orientation distributions based on diffusion magnetic resonance imaging is a critical step in white matter tractography imaging. However, due to the complex structure of superficial white matter, existing models for estimating fiber orientation distributions are ineffective in reconstructing superficial white matter and even reconstruct incorrect orientation distributions.</div></div><div><h3>New method</h3><div>In this paper, we improve the traditional constrained spherical deconvolution method and propose a novel asymmetric constrained spherical deconvolution method. The method takes into account that the displacement profile of the water molecules in brain tissue are non-Gaussian diffusion and the core parameter kurtosis might characterize tissue structure better than diffusivity coefficients. So diffusion kurtosis imaging model is used to estimate the white matter response function. The proposed method applies the diffusion kurtosis imaging model response function to the asymmetric fiber orientation distributions, and this is the first attempt to obtain more accurate fiber orientation distributions. Furthermore, the Gaussian-Distribution distance weight and Watson-Distribution angle weight are used for asymmetric regularization.</div></div><div><h3>Results</h3><div>We evaluate the method using FiberCup phantom, ISMRM 2015 data and in vivo data provided CHCP dataset. The results show that our proposed method can more accurately reconstruct the complex fiber structure of superficial white matter with more accurate fiber orientation, fewer pseudo-peaks, and mitigate gyral bias.</div></div><div><h3>Comparison with existing methods</h3><div>Our proposed method has higher accuracy in estimating the fiber orientation distributions and can reconstruct highly curved fiber voxels.</div></div><div><h3>Conclusion</h3><div>This proposed method provides new insights into the estimation of the orientation distribution of superficial white matter fibers.</div></div>","PeriodicalId":16415,"journal":{"name":"Journal of Neuroscience Methods","volume":"415 ","pages":"Article 110353"},"PeriodicalIF":2.7,"publicationDate":"2024-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143177173","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Periodicity-based multi-dimensional interaction convolution network with multi-scale feature fusion for motor imagery EEG classification

IF 2.7 4区医学 Q2 BIOCHEMICAL RESEARCH METHODS

Journal of Neuroscience Methods

Pub Date : 2024-12-25 DOI: 10.1016/j.jneumeth.2024.110356

Yunshuo Dai, Xiao Deng, Xiuli Fu, Yixin Zhao

Background

The Motor Imagery (MI)-based Brain-Computer Interface (BCI) has vast potential in fields such as medical rehabilitation and control engineering. In recent years, MI decoding methods based on deep learning have gained extensive attention. However, capturing the complex dynamic changes in EEG signals remains a challenge, and the decoding performance still needs further improvement.

New methods

The paper proposes a novel method, Periodicity-based Multi-Dimensional Interaction Convolution Network with Multi-Scale Feature Fusion (PMD-MSNet), for MI-EEG signal classification. It converts 1D EEG signals into multi-period 2D tensors to capture intra-period and inter-period variations and enables cross-dimensional interaction based on periodic features. Subsequently, parallel multi-scale convolution is utilized to adaptively extract temporal, frequency, and time-frequency features.

Results

Experimental results on the BCI IV-2a dataset demonstrate that the PMD-MSNet model achieves a classification accuracy of 82.25 % on average and a kappa value of 0.763, which significantly outperforms seven other deep learning-based EEG decoding models. The model attained the highest classification accuracy and kappa value among the seven subjects, showcasing its superior performance and robustness.

Conclusions

The PMD-MSNet model incorporates periodic features, multi-dimensional interaction mechanisms, multi-scale convolutions to achieve efficient feature extraction and classification of EEG signals, significantly enhancing the performance of MI classification tasks.

{"title":"Periodicity-based multi-dimensional interaction convolution network with multi-scale feature fusion for motor imagery EEG classification","authors":"Yunshuo Dai, Xiao Deng, Xiuli Fu, Yixin Zhao","doi":"10.1016/j.jneumeth.2024.110356","DOIUrl":"10.1016/j.jneumeth.2024.110356","url":null,"abstract":"<div><h3>Background</h3><div>The Motor Imagery (MI)-based Brain-Computer Interface (BCI) has vast potential in fields such as medical rehabilitation and control engineering. In recent years, MI decoding methods based on deep learning have gained extensive attention. However, capturing the complex dynamic changes in EEG signals remains a challenge, and the decoding performance still needs further improvement.</div></div><div><h3>New methods</h3><div>The paper proposes a novel method, Periodicity-based Multi-Dimensional Interaction Convolution Network with Multi-Scale Feature Fusion (PMD-MSNet), for MI-EEG signal classification. It converts 1D EEG signals into multi-period 2D tensors to capture intra-period and inter-period variations and enables cross-dimensional interaction based on periodic features. Subsequently, parallel multi-scale convolution is utilized to adaptively extract temporal, frequency, and time-frequency features.</div></div><div><h3>Results</h3><div>Experimental results on the BCI IV-2a dataset demonstrate that the PMD-MSNet model achieves a classification accuracy of 82.25 % on average and a kappa value of 0.763, which significantly outperforms seven other deep learning-based EEG decoding models. The model attained the highest classification accuracy and kappa value among the seven subjects, showcasing its superior performance and robustness.</div></div><div><h3>Conclusions</h3><div>The PMD-MSNet model incorporates periodic features, multi-dimensional interaction mechanisms, multi-scale convolutions to achieve efficient feature extraction and classification of EEG signals, significantly enhancing the performance of MI classification tasks.</div></div>","PeriodicalId":16415,"journal":{"name":"Journal of Neuroscience Methods","volume":"415 ","pages":"Article 110356"},"PeriodicalIF":2.7,"publicationDate":"2024-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143176075","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0