Electroencephalogram (EEG) plays a pivotal role in the detection and analysis of epileptic seizures, which affects over 70 million people in the world. Nonetheless, the visual interpretation of EEG signals for epilepsy detection is laborious and time-consuming. To tackle this open challenge, we introduce a straightforward yet efficient hybrid deep learning approach, named ResBiLSTM, for detecting epileptic seizures using EEG signals. Firstly, a one-dimensional residual neural network (ResNet) is tailored to adeptly extract the local spatial features of EEG signals. Subsequently, the acquired features are input into a bidirectional long short-term memory (BiLSTM) layer to model temporal dependencies. These output features are further processed through two fully connected layers to achieve the final epileptic seizure detection. The performance of ResBiLSTM is assessed on the epileptic seizure datasets provided by the University of Bonn and Temple University Hospital (TUH). The ResBiLSTM model achieves epileptic seizure detection accuracy rates of 98.88-100% in binary and ternary classifications on the Bonn dataset. Experimental outcomes for seizure recognition across seven epilepsy seizure types on the TUH seizure corpus (TUSZ) dataset indicate that the ResBiLSTM model attains a classification accuracy of 95.03% and a weighted F1 score of 95.03% with 10-fold cross-validation. These findings illustrate that ResBiLSTM outperforms several recent deep learning state-of-the-art approaches.
{"title":"Residual and bidirectional LSTM for epileptic seizure detection.","authors":"Wei Zhao, Wen-Feng Wang, Lalit Mohan Patnaik, Bao-Can Zhang, Su-Jun Weng, Shi-Xiao Xiao, De-Zhi Wei, Hai-Feng Zhou","doi":"10.3389/fncom.2024.1415967","DOIUrl":"10.3389/fncom.2024.1415967","url":null,"abstract":"<p><p>Electroencephalogram (EEG) plays a pivotal role in the detection and analysis of epileptic seizures, which affects over 70 million people in the world. Nonetheless, the visual interpretation of EEG signals for epilepsy detection is laborious and time-consuming. To tackle this open challenge, we introduce a straightforward yet efficient hybrid deep learning approach, named ResBiLSTM, for detecting epileptic seizures using EEG signals. Firstly, a one-dimensional residual neural network (ResNet) is tailored to adeptly extract the local spatial features of EEG signals. Subsequently, the acquired features are input into a bidirectional long short-term memory (BiLSTM) layer to model temporal dependencies. These output features are further processed through two fully connected layers to achieve the final epileptic seizure detection. The performance of ResBiLSTM is assessed on the epileptic seizure datasets provided by the University of Bonn and Temple University Hospital (TUH). The ResBiLSTM model achieves epileptic seizure detection accuracy rates of 98.88-100% in binary and ternary classifications on the Bonn dataset. Experimental outcomes for seizure recognition across seven epilepsy seizure types on the TUH seizure corpus (TUSZ) dataset indicate that the ResBiLSTM model attains a classification accuracy of 95.03% and a weighted F1 score of 95.03% with 10-fold cross-validation. These findings illustrate that ResBiLSTM outperforms several recent deep learning state-of-the-art approaches.</p>","PeriodicalId":12363,"journal":{"name":"Frontiers in Computational Neuroscience","volume":null,"pages":null},"PeriodicalIF":2.1,"publicationDate":"2024-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11215953/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141476309","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}
Pub Date : 2024-06-14DOI: 10.3389/fncom.2024.1352685
Rousslan Fernand Julien Dossa, Kai Arulkumaran, Arthur Juliani, Shuntaro Sasai, Ryota Kanai
As the apparent intelligence of artificial neural networks (ANNs) advances, they are increasingly likened to the functional networks and information processing capabilities of the human brain. Such comparisons have typically focused on particular modalities, such as vision or language. The next frontier is to use the latest advances in ANNs to design and investigate scalable models of higher-level cognitive processes, such as conscious information access, which have historically lacked concrete and specific hypotheses for scientific evaluation. In this work, we propose and then empirically assess an embodied agent with a structure based on global workspace theory (GWT) as specified in the recently proposed “indicator properties” of consciousness. In contrast to prior works on GWT which utilized single modalities, our agent is trained to navigate 3D environments based on realistic audiovisual inputs. We find that the global workspace architecture performs better and more robustly at smaller working memory sizes, as compared to a standard recurrent architecture. Beyond performance, we perform a series of analyses on the learned representations of our architecture and share findings that point to task complexity and regularization being essential for feature learning and the development of meaningful attentional patterns within the workspace.
{"title":"Design and evaluation of a global workspace agent embodied in a realistic multimodal environment","authors":"Rousslan Fernand Julien Dossa, Kai Arulkumaran, Arthur Juliani, Shuntaro Sasai, Ryota Kanai","doi":"10.3389/fncom.2024.1352685","DOIUrl":"https://doi.org/10.3389/fncom.2024.1352685","url":null,"abstract":"As the apparent intelligence of artificial neural networks (ANNs) advances, they are increasingly likened to the functional networks and information processing capabilities of the human brain. Such comparisons have typically focused on particular modalities, such as vision or language. The next frontier is to use the latest advances in ANNs to design and investigate scalable models of higher-level cognitive processes, such as conscious information access, which have historically lacked concrete and specific hypotheses for scientific evaluation. In this work, we propose and then empirically assess an embodied agent with a structure based on global workspace theory (GWT) as specified in the recently proposed “indicator properties” of consciousness. In contrast to prior works on GWT which utilized single modalities, our agent is trained to navigate 3D environments based on realistic audiovisual inputs. We find that the global workspace architecture performs better and more robustly at smaller working memory sizes, as compared to a standard recurrent architecture. Beyond performance, we perform a series of analyses on the learned representations of our architecture and share findings that point to task complexity and regularization being essential for feature learning and the development of meaningful attentional patterns within the workspace.","PeriodicalId":12363,"journal":{"name":"Frontiers in Computational Neuroscience","volume":null,"pages":null},"PeriodicalIF":3.2,"publicationDate":"2024-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141339724","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}
Pub Date : 2024-06-12DOI: 10.3389/fncom.2024.1414462
Tae Hoon Kim, Moez Krichen, Stephen Ojo, Gabriel Avelino R. Sampedro, Meznah A. Alamro
Parkinson's disease (PD) is a globally significant health challenge, necessitating accurate and timely diagnostic methods to facilitate effective treatment and intervention. In recent years, self-supervised deep representation pattern learning (SS-DRPL) has emerged as a promising approach for extracting valuable representations from data, offering the potential to enhance the efficiency of voice-based PD detection. This research study focuses on investigating the utilization of SS-DRPL in conjunction with deep learning algorithms for voice-based PD classification. This study encompasses a comprehensive evaluation aimed at assessing the accuracy of various predictive models, particularly deep learning methods when combined with SS-DRPL. Two deep learning architectures, namely hybrid Long Short-Term Memory and Recurrent Neural Networks (LSTM-RNN) and Deep Neural Networks (DNN), are employed and compared in terms of their ability to detect voice-based PD cases accurately. Additionally, several traditional machine learning models are also included to establish a baseline for comparison. The findings of the study reveal that the incorporation of SS-DRPL leads to improved model performance across all experimental setups. Notably, the LSTM-RNN architecture augmented with SS-DRPL achieves the highest F1-score of 0.94, indicating its superior ability to detect PD cases using voice-based data effectively. This outcome underscores the efficacy of SS-DRPL in enabling deep learning models to learn intricate patterns and correlations within the data, thereby facilitating more accurate PD classification.
{"title":"SS-DRPL: self-supervised deep representation pattern learning for voice-based Parkinson's disease detection","authors":"Tae Hoon Kim, Moez Krichen, Stephen Ojo, Gabriel Avelino R. Sampedro, Meznah A. Alamro","doi":"10.3389/fncom.2024.1414462","DOIUrl":"https://doi.org/10.3389/fncom.2024.1414462","url":null,"abstract":"Parkinson's disease (PD) is a globally significant health challenge, necessitating accurate and timely diagnostic methods to facilitate effective treatment and intervention. In recent years, self-supervised deep representation pattern learning (SS-DRPL) has emerged as a promising approach for extracting valuable representations from data, offering the potential to enhance the efficiency of voice-based PD detection. This research study focuses on investigating the utilization of SS-DRPL in conjunction with deep learning algorithms for voice-based PD classification. This study encompasses a comprehensive evaluation aimed at assessing the accuracy of various predictive models, particularly deep learning methods when combined with SS-DRPL. Two deep learning architectures, namely hybrid Long Short-Term Memory and Recurrent Neural Networks (LSTM-RNN) and Deep Neural Networks (DNN), are employed and compared in terms of their ability to detect voice-based PD cases accurately. Additionally, several traditional machine learning models are also included to establish a baseline for comparison. The findings of the study reveal that the incorporation of SS-DRPL leads to improved model performance across all experimental setups. Notably, the LSTM-RNN architecture augmented with SS-DRPL achieves the highest F1-score of 0.94, indicating its superior ability to detect PD cases using voice-based data effectively. This outcome underscores the efficacy of SS-DRPL in enabling deep learning models to learn intricate patterns and correlations within the data, thereby facilitating more accurate PD classification.","PeriodicalId":12363,"journal":{"name":"Frontiers in Computational Neuroscience","volume":null,"pages":null},"PeriodicalIF":3.2,"publicationDate":"2024-06-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141352053","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}
Pub Date : 2024-06-12DOI: 10.3389/fncom.2024.1418546
Eid Albalawi, Arastu Thakur, D. Dorai, Surbhi Bhatia Khan, T. Mahesh, Ahlam Almusharraf, Khursheed Aurangzeb, Muhammad Shahid Anwar
The necessity of prompt and accurate brain tumor diagnosis is unquestionable for optimizing treatment strategies and patient prognoses. Traditional reliance on Magnetic Resonance Imaging (MRI) analysis, contingent upon expert interpretation, grapples with challenges such as time-intensive processes and susceptibility to human error.This research presents a novel Convolutional Neural Network (CNN) architecture designed to enhance the accuracy and efficiency of brain tumor detection in MRI scans.The dataset used in the study comprises 7,023 brain MRI images from figshare, SARTAJ, and Br35H, categorized into glioma, meningioma, no tumor, and pituitary classes, with a CNN-based multi-task classification model employed for tumor detection, classification, and location identification. Our methodology focused on multi-task classification using a single CNN model for various brain MRI classification tasks, including tumor detection, classification based on grade and type, and tumor location identification.The proposed CNN model incorporates advanced feature extraction capabilities and deep learning optimization techniques, culminating in a groundbreaking paradigm shift in automated brain MRI analysis. With an exceptional tumor classification accuracy of 99%, our method surpasses current methodologies, demonstrating the remarkable potential of deep learning in medical applications.This study represents a significant advancement in the early detection and treatment planning of brain tumors, offering a more efficient and accurate alternative to traditional MRI analysis methods.
{"title":"Enhancing brain tumor classification in MRI scans with a multi-layer customized convolutional neural network approach","authors":"Eid Albalawi, Arastu Thakur, D. Dorai, Surbhi Bhatia Khan, T. Mahesh, Ahlam Almusharraf, Khursheed Aurangzeb, Muhammad Shahid Anwar","doi":"10.3389/fncom.2024.1418546","DOIUrl":"https://doi.org/10.3389/fncom.2024.1418546","url":null,"abstract":"The necessity of prompt and accurate brain tumor diagnosis is unquestionable for optimizing treatment strategies and patient prognoses. Traditional reliance on Magnetic Resonance Imaging (MRI) analysis, contingent upon expert interpretation, grapples with challenges such as time-intensive processes and susceptibility to human error.This research presents a novel Convolutional Neural Network (CNN) architecture designed to enhance the accuracy and efficiency of brain tumor detection in MRI scans.The dataset used in the study comprises 7,023 brain MRI images from figshare, SARTAJ, and Br35H, categorized into glioma, meningioma, no tumor, and pituitary classes, with a CNN-based multi-task classification model employed for tumor detection, classification, and location identification. Our methodology focused on multi-task classification using a single CNN model for various brain MRI classification tasks, including tumor detection, classification based on grade and type, and tumor location identification.The proposed CNN model incorporates advanced feature extraction capabilities and deep learning optimization techniques, culminating in a groundbreaking paradigm shift in automated brain MRI analysis. With an exceptional tumor classification accuracy of 99%, our method surpasses current methodologies, demonstrating the remarkable potential of deep learning in medical applications.This study represents a significant advancement in the early detection and treatment planning of brain tumors, offering a more efficient and accurate alternative to traditional MRI analysis methods.","PeriodicalId":12363,"journal":{"name":"Frontiers in Computational Neuroscience","volume":null,"pages":null},"PeriodicalIF":3.2,"publicationDate":"2024-06-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141353667","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}
Pub Date : 2024-06-06DOI: 10.3389/fncom.2024.1410335
Astrid Correa, Adam Ponzi, Vladimir M. Calderón, Rosanna Migliore
Under normal conditions the principal cells of the striatum, medium spiny neurons (MSNs), show structured cell assembly activity patterns which alternate sequentially over exceedingly long timescales of many minutes. It is important to understand this activity since it is characteristically disrupted in multiple pathologies, such as Parkinson's disease and dyskinesia, and thought to be caused by alterations in the MSN to MSN lateral inhibitory connections and in the strength and distribution of cortical excitation to MSNs. To understand how these long timescales arise we extended a previous network model of MSN cells to include synapses with short-term plasticity, with parameters taken from a recent detailed striatal connectome study. We first confirmed the presence of sequentially switching cell clusters using the non-linear dimensionality reduction technique, Uniform Manifold Approximation and Projection (UMAP). We found that the network could generate non-stationary activity patterns varying extremely slowly on the order of minutes under biologically realistic conditions. Next we used Simulation Based Inference (SBI) to train a deep net to map features of the MSN network generated cell assembly activity to MSN network parameters. We used the trained SBI model to estimate MSN network parameters from ex-vivo brain slice calcium imaging data. We found that best fit network parameters were very close to their physiologically observed values. On the other hand network parameters estimated from Parkinsonian, decorticated and dyskinetic ex-vivo slice preparations were different. Our work may provide a pipeline for diagnosis of basal ganglia pathology from spiking data as well as for the design pharmacological treatments.
{"title":"Pathological cell assembly dynamics in a striatal MSN network model","authors":"Astrid Correa, Adam Ponzi, Vladimir M. Calderón, Rosanna Migliore","doi":"10.3389/fncom.2024.1410335","DOIUrl":"https://doi.org/10.3389/fncom.2024.1410335","url":null,"abstract":"Under normal conditions the principal cells of the striatum, medium spiny neurons (MSNs), show structured cell assembly activity patterns which alternate sequentially over exceedingly long timescales of many minutes. It is important to understand this activity since it is characteristically disrupted in multiple pathologies, such as Parkinson's disease and dyskinesia, and thought to be caused by alterations in the MSN to MSN lateral inhibitory connections and in the strength and distribution of cortical excitation to MSNs. To understand how these long timescales arise we extended a previous network model of MSN cells to include synapses with short-term plasticity, with parameters taken from a recent detailed striatal connectome study. We first confirmed the presence of sequentially switching cell clusters using the non-linear dimensionality reduction technique, Uniform Manifold Approximation and Projection (UMAP). We found that the network could generate non-stationary activity patterns varying extremely slowly on the order of minutes under biologically realistic conditions. Next we used Simulation Based Inference (SBI) to train a deep net to map features of the MSN network generated cell assembly activity to MSN network parameters. We used the trained SBI model to estimate MSN network parameters from ex-vivo brain slice calcium imaging data. We found that best fit network parameters were very close to their physiologically observed values. On the other hand network parameters estimated from Parkinsonian, decorticated and dyskinetic ex-vivo slice preparations were different. Our work may provide a pipeline for diagnosis of basal ganglia pathology from spiking data as well as for the design pharmacological treatments.","PeriodicalId":12363,"journal":{"name":"Frontiers in Computational Neuroscience","volume":null,"pages":null},"PeriodicalIF":3.2,"publicationDate":"2024-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141376788","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}
Pub Date : 2024-05-30DOI: 10.3389/fncom.2024.1355855
Akito Fukunishi, Kyo Kutsuzawa, Dai Owaki, Mitsuhiro Hayashibe
How our central nervous system efficiently controls our complex musculoskeletal system is still debated. The muscle synergy hypothesis is proposed to simplify this complex system by assuming the existence of functional neural modules that coordinate several muscles. Modularity based on muscle synergies can facilitate motor learning without compromising task performance. However, the effectiveness of modularity in motor control remains debated. This ambiguity can, in part, stem from overlooking that the performance of modularity depends on the mechanical aspects of modules of interest, such as the torque the modules exert. To address this issue, this study introduces two criteria to evaluate the quality of module sets based on commonly used performance metrics in motor learning studies: the accuracy of torque production and learning speed. One evaluates the regularity in the direction of mechanical torque the modules exert, while the other evaluates the evenness of its magnitude. For verification of our criteria, we simulated motor learning of torque production tasks in a realistic musculoskeletal system of the upper arm using feed-forward neural networks while changing the control conditions. We found that the proposed criteria successfully explain the tendency of learning performance in various control conditions. These result suggest that regularity in the direction of and evenness in magnitude of mechanical torque of utilized modules are significant factor for determining learning performance. Although the criteria were originally conceived for an error-based learning scheme, the approach to pursue which set of modules is better for motor control can have significant implications in other studies of modularity in general.
{"title":"Synergy quality assessment of muscle modules for determining learning performance using a realistic musculoskeletal model","authors":"Akito Fukunishi, Kyo Kutsuzawa, Dai Owaki, Mitsuhiro Hayashibe","doi":"10.3389/fncom.2024.1355855","DOIUrl":"https://doi.org/10.3389/fncom.2024.1355855","url":null,"abstract":"How our central nervous system efficiently controls our complex musculoskeletal system is still debated. The muscle synergy hypothesis is proposed to simplify this complex system by assuming the existence of functional neural modules that coordinate several muscles. Modularity based on muscle synergies can facilitate motor learning without compromising task performance. However, the effectiveness of modularity in motor control remains debated. This ambiguity can, in part, stem from overlooking that the performance of modularity depends on the mechanical aspects of modules of interest, such as the torque the modules exert. To address this issue, this study introduces two criteria to evaluate the quality of module sets based on commonly used performance metrics in motor learning studies: the accuracy of torque production and learning speed. One evaluates the regularity in the direction of mechanical torque the modules exert, while the other evaluates the evenness of its magnitude. For verification of our criteria, we simulated motor learning of torque production tasks in a realistic musculoskeletal system of the upper arm using feed-forward neural networks while changing the control conditions. We found that the proposed criteria successfully explain the tendency of learning performance in various control conditions. These result suggest that regularity in the direction of and evenness in magnitude of mechanical torque of utilized modules are significant factor for determining learning performance. Although the criteria were originally conceived for an error-based learning scheme, the approach to pursue which set of modules is better for motor control can have significant implications in other studies of modularity in general.","PeriodicalId":12363,"journal":{"name":"Frontiers in Computational Neuroscience","volume":null,"pages":null},"PeriodicalIF":3.2,"publicationDate":"2024-05-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141189235","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}
The spiking convolutional neural network (SCNN) is a kind of spiking neural network (SNN) with high accuracy for visual tasks and power efficiency on neuromorphic hardware, which is attractive for edge applications. However, it is challenging to implement SCNNs on resource-constrained edge devices because of the large number of convolutional operations and membrane potential (Vm) storage needed. Previous works have focused on timestep reduction, network pruning, and network quantization to realize SCNN implementation on edge devices. However, they overlooked similarities between spiking feature maps (SFmaps), which contain significant redundancy and cause unnecessary computation and storage. This work proposes a dual-threshold spiking convolutional neural network (DT-SCNN) to decrease the number of operations and memory access by utilizing similarities between SFmaps. The DT-SCNN employs dual firing thresholds to derive two similar SFmaps from one Vm map, reducing the number of convolutional operations and decreasing the volume of Vms and convolutional weights by half. We propose a variant spatio-temporal back propagation (STBP) training method with a two-stage strategy to train DT-SCNNs to decrease the inference timestep to 1. The experimental results show that the dual-thresholds mechanism achieves a 50% reduction in operations and data storage for the convolutional layers compared to conventional SCNNs while achieving not more than a 0.4% accuracy loss on the CIFAR10, MNIST, and Fashion MNIST datasets. Due to the lightweight network and single timestep inference, the DT-SCNN has the least number of operations compared to previous works, paving the way for low-latency and power-efficient edge applications.
尖峰卷积神经网络(SCNN)是一种尖峰神经网络(SNN),在视觉任务中具有高精确度,在神经形态硬件上具有高能效,对边缘应用很有吸引力。然而,由于需要大量卷积运算和膜电位(Vm)存储,在资源受限的边缘设备上实现 SCNN 是一项挑战。以前的工作主要集中在减少时间步长、网络剪枝和网络量化上,以实现在边缘设备上实施 SCNN。然而,他们忽略了尖峰特征图(SFmaps)之间的相似性,这些特征图包含大量冗余,会造成不必要的计算和存储。本研究提出了一种双阈值尖峰卷积神经网络(DT-SCNN),通过利用 SFmaps 之间的相似性来减少运算次数和内存访问。DT-SCNN 采用双发射阈值,从一个 Vm 映射中推导出两个相似的 SF 映射,从而减少了卷积操作的数量,并将 Vm 和卷积权重的体积减少了一半。实验结果表明,与传统 SCNN 相比,双阈值机制减少了卷积层 50% 的操作和数据存储,同时在 CIFAR10、MNIST 和时尚 MNIST 数据集上的准确率损失不超过 0.4%。由于采用了轻量级网络和单时间步推理,DT-SCNN 的操作次数与以前的作品相比最少,为低延迟、高能效的边缘应用铺平了道路。
{"title":"DT-SCNN: dual-threshold spiking convolutional neural network with fewer operations and memory access for edge applications","authors":"Fuming Lei, Xu Yang, Jian Liu, Runjiang Dou, Nanjian Wu","doi":"10.3389/fncom.2024.1418115","DOIUrl":"https://doi.org/10.3389/fncom.2024.1418115","url":null,"abstract":"The spiking convolutional neural network (SCNN) is a kind of spiking neural network (SNN) with high accuracy for visual tasks and power efficiency on neuromorphic hardware, which is attractive for edge applications. However, it is challenging to implement SCNNs on resource-constrained edge devices because of the large number of convolutional operations and membrane potential (Vm) storage needed. Previous works have focused on timestep reduction, network pruning, and network quantization to realize SCNN implementation on edge devices. However, they overlooked similarities between spiking feature maps (SFmaps), which contain significant redundancy and cause unnecessary computation and storage. This work proposes a dual-threshold spiking convolutional neural network (DT-SCNN) to decrease the number of operations and memory access by utilizing similarities between SFmaps. The DT-SCNN employs dual firing thresholds to derive two similar SFmaps from one Vm map, reducing the number of convolutional operations and decreasing the volume of Vms and convolutional weights by half. We propose a variant spatio-temporal back propagation (STBP) training method with a two-stage strategy to train DT-SCNNs to decrease the inference timestep to 1. The experimental results show that the dual-thresholds mechanism achieves a 50% reduction in operations and data storage for the convolutional layers compared to conventional SCNNs while achieving not more than a 0.4% accuracy loss on the CIFAR10, MNIST, and Fashion MNIST datasets. Due to the lightweight network and single timestep inference, the DT-SCNN has the least number of operations compared to previous works, paving the way for low-latency and power-efficient edge applications.","PeriodicalId":12363,"journal":{"name":"Frontiers in Computational Neuroscience","volume":null,"pages":null},"PeriodicalIF":3.2,"publicationDate":"2024-05-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141189272","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}
Pub Date : 2024-05-28DOI: 10.3389/fncom.2024.1398898
Bernard A. Pailthorpe
Network analysis of the marmoset cortical connectivity data indicates a significant 3D cluster in and around the pre-frontal cortex. A multi-node, heterogeneous neural mass model of this six-node cluster was constructed. Its parameters were informed by available experimental and simulation data so that each neural mass oscillated in a characteristic frequency band. Nodes were connected with directed, weighted links derived from the marmoset structural connectivity data. Heterogeneity arose from the different link weights and model parameters for each node. Stimulation of the cluster with an incident pulse train modulated in the standard frequency bands induced a variety of dynamical state transitions that lasted in the range of 5–10 s, suggestive of timescales relevant to short-term memory. A short gamma burst rapidly reset the beta-induced transition. The theta-induced transition state showed a spontaneous, delayed reset to the resting state. An additional, continuous gamma wave stimulus induced a new beating oscillatory state. Longer or repeated gamma bursts were phase-aligned with the beta oscillation, delivering increasing energy input and causing shorter transition times. The relevance of these results to working memory is yet to be established, but they suggest interesting opportunities.
{"title":"Simulated dynamical transitions in a heterogeneous marmoset pFC cluster","authors":"Bernard A. Pailthorpe","doi":"10.3389/fncom.2024.1398898","DOIUrl":"https://doi.org/10.3389/fncom.2024.1398898","url":null,"abstract":"Network analysis of the marmoset cortical connectivity data indicates a significant 3D cluster in and around the pre-frontal cortex. A multi-node, heterogeneous neural mass model of this six-node cluster was constructed. Its parameters were informed by available experimental and simulation data so that each neural mass oscillated in a characteristic frequency band. Nodes were connected with directed, weighted links derived from the marmoset structural connectivity data. Heterogeneity arose from the different link weights and model parameters for each node. Stimulation of the cluster with an incident pulse train modulated in the standard frequency bands induced a variety of dynamical state transitions that lasted in the range of 5–10 s, suggestive of timescales relevant to short-term memory. A short gamma burst rapidly reset the beta-induced transition. The theta-induced transition state showed a spontaneous, delayed reset to the resting state. An additional, continuous gamma wave stimulus induced a new beating oscillatory state. Longer or repeated gamma bursts were phase-aligned with the beta oscillation, delivering increasing energy input and causing shorter transition times. The relevance of these results to working memory is yet to be established, but they suggest interesting opportunities.","PeriodicalId":12363,"journal":{"name":"Frontiers in Computational Neuroscience","volume":null,"pages":null},"PeriodicalIF":3.2,"publicationDate":"2024-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141167607","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}
MRI is one of the commonly used diagnostic methods in clinical practice, especially in brain diseases. There are many sequences in MRI, but T1CE images can only be obtained by using contrast agents. Many patients (such as cancer patients) must undergo alignment of multiple MRI sequences for diagnosis, especially the contrast-enhanced magnetic resonance sequence. However, some patients such as pregnant women, children, etc. find it difficult to use contrast agents to obtain enhanced sequences, and contrast agents have many adverse reactions, which can pose a significant risk. With the continuous development of deep learning, the emergence of generative adversarial networks makes it possible to extract features from one type of image to generate another type of image.We propose a generative adversarial network model with multimodal inputs and end-to-end decoding based on the pix2pix model. For the pix2pix model, we used four evaluation metrics: NMSE, RMSE, SSIM, and PNSR to assess the effectiveness of our generated model.Through statistical analysis, we compared our proposed new model with pix2pix and found significant differences between the two. Our model outperformed pix2pix, with higher SSIM and PNSR, lower NMSE and RMSE. We also found that the input of T1W images and T2W images had better effects than other combinations, providing new ideas for subsequent work on generating magnetic resonance enhancement sequence images. By using our model, it is possible to generate magnetic resonance enhanced sequence images based on magnetic resonance non-enhanced sequence images.This has significant implications as it can greatly reduce the use of contrast agents to protect populations such as pregnant women and children who are contraindicated for contrast agents. Additionally, contrast agents are relatively expensive, and this generation method may bring about substantial economic benefits.
{"title":"Multi-sequence generative adversarial network: better generation for enhanced magnetic resonance imaging images","authors":"Leizi Li, Jingchun Yu, Yijin Li, Jinbo Wei, Ruifang Fan, Dieen Wu, Yufeng Ye","doi":"10.3389/fncom.2024.1365238","DOIUrl":"https://doi.org/10.3389/fncom.2024.1365238","url":null,"abstract":"MRI is one of the commonly used diagnostic methods in clinical practice, especially in brain diseases. There are many sequences in MRI, but T1CE images can only be obtained by using contrast agents. Many patients (such as cancer patients) must undergo alignment of multiple MRI sequences for diagnosis, especially the contrast-enhanced magnetic resonance sequence. However, some patients such as pregnant women, children, etc. find it difficult to use contrast agents to obtain enhanced sequences, and contrast agents have many adverse reactions, which can pose a significant risk. With the continuous development of deep learning, the emergence of generative adversarial networks makes it possible to extract features from one type of image to generate another type of image.We propose a generative adversarial network model with multimodal inputs and end-to-end decoding based on the pix2pix model. For the pix2pix model, we used four evaluation metrics: NMSE, RMSE, SSIM, and PNSR to assess the effectiveness of our generated model.Through statistical analysis, we compared our proposed new model with pix2pix and found significant differences between the two. Our model outperformed pix2pix, with higher SSIM and PNSR, lower NMSE and RMSE. We also found that the input of T1W images and T2W images had better effects than other combinations, providing new ideas for subsequent work on generating magnetic resonance enhancement sequence images. By using our model, it is possible to generate magnetic resonance enhanced sequence images based on magnetic resonance non-enhanced sequence images.This has significant implications as it can greatly reduce the use of contrast agents to protect populations such as pregnant women and children who are contraindicated for contrast agents. Additionally, contrast agents are relatively expensive, and this generation method may bring about substantial economic benefits.","PeriodicalId":12363,"journal":{"name":"Frontiers in Computational Neuroscience","volume":null,"pages":null},"PeriodicalIF":3.2,"publicationDate":"2024-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141110229","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}
Pub Date : 2024-05-22DOI: 10.3389/fncom.2024.1392655
Andrew S. Perley, Todd P. Coleman
Cross frequency coupling (CFC) between electrophysiological signals in the brain is a long-studied phenomenon and its abnormalities have been observed in conditions such as Parkinson's disease and epilepsy. More recently, CFC has been observed in stomach-brain electrophysiologic studies and thus becomes an enticing possible target for diseases involving aberrations of the gut-brain axis. However, current methods of detecting coupling, specifically phase-amplitude coupling (PAC), do not attempt to capture the phase and amplitude statistical relationships.In this paper, we first demonstrate a method of modeling these joint statistics with a flexible parametric approach, where we model the conditional distribution of amplitude given phase using a gamma distributed generalized linear model (GLM) with a Fourier basis of regressors. We perform model selection with minimum description length (MDL) principle, demonstrate a method for assessing goodness-of-fit (GOF), and showcase the efficacy of this approach in multiple electroencephalography (EEG) datasets. Secondly, we showcase how we can utilize the mutual information, which operates on the joint distribution, as a canonical measure of coupling, as it is non-zero and non-negative if and only if the phase and amplitude are not statistically independent. In addition, we build off of previous work by Martinez-Cancino et al., and Voytek et al., and show that the information density, evaluated using our method along the given sample path, is a promising measure of time-resolved PAC.Using synthetically generated gut-brain coupled signals, we demonstrate that our method outperforms the existing gold-standard methods for detectable low-levels of phase-amplitude coupling through receiver operating characteristic (ROC) curve analysis. To validate our method, we test on invasive EEG recordings by generating comodulograms, and compare our method to the gold standard PAC measure, Modulation Index, demonstrating comparable performance in exploratory analysis. Furthermore, to showcase its use in joint gut-brain electrophysiology data, we generate topoplots of simultaneous high-density EEG and electrgastrography recordings and reproduce seminal work by Richter et al. that demonstrated the existence of gut-brain PAC. Using simulated data, we validate our method for different types of time-varying coupling and then demonstrate its performance to track time-varying PAC in sleep spindle EEG and mismatch negativity (MMN) datasets.Our new measure of PAC using Gamma GLMs and mutual information demonstrates a promising new way to compute PAC values using the full joint distribution on amplitude and phase. Our measure outperforms the most common existing measures of PAC, and show promising results in identifying time varying PAC in electrophysiological datasets. In addition, we provide for using our method with multiple comparisons and show that our measure potentially has more statistical power in electrophysiologic recordin
{"title":"A mutual information measure of phase-amplitude coupling using gamma generalized linear models","authors":"Andrew S. Perley, Todd P. Coleman","doi":"10.3389/fncom.2024.1392655","DOIUrl":"https://doi.org/10.3389/fncom.2024.1392655","url":null,"abstract":"Cross frequency coupling (CFC) between electrophysiological signals in the brain is a long-studied phenomenon and its abnormalities have been observed in conditions such as Parkinson's disease and epilepsy. More recently, CFC has been observed in stomach-brain electrophysiologic studies and thus becomes an enticing possible target for diseases involving aberrations of the gut-brain axis. However, current methods of detecting coupling, specifically phase-amplitude coupling (PAC), do not attempt to capture the phase and amplitude statistical relationships.In this paper, we first demonstrate a method of modeling these joint statistics with a flexible parametric approach, where we model the conditional distribution of amplitude given phase using a gamma distributed generalized linear model (GLM) with a Fourier basis of regressors. We perform model selection with minimum description length (MDL) principle, demonstrate a method for assessing goodness-of-fit (GOF), and showcase the efficacy of this approach in multiple electroencephalography (EEG) datasets. Secondly, we showcase how we can utilize the mutual information, which operates on the joint distribution, as a canonical measure of coupling, as it is non-zero and non-negative if and only if the phase and amplitude are not statistically independent. In addition, we build off of previous work by Martinez-Cancino et al., and Voytek et al., and show that the information density, evaluated using our method along the given sample path, is a promising measure of time-resolved PAC.Using synthetically generated gut-brain coupled signals, we demonstrate that our method outperforms the existing gold-standard methods for detectable low-levels of phase-amplitude coupling through receiver operating characteristic (ROC) curve analysis. To validate our method, we test on invasive EEG recordings by generating comodulograms, and compare our method to the gold standard PAC measure, Modulation Index, demonstrating comparable performance in exploratory analysis. Furthermore, to showcase its use in joint gut-brain electrophysiology data, we generate topoplots of simultaneous high-density EEG and electrgastrography recordings and reproduce seminal work by Richter et al. that demonstrated the existence of gut-brain PAC. Using simulated data, we validate our method for different types of time-varying coupling and then demonstrate its performance to track time-varying PAC in sleep spindle EEG and mismatch negativity (MMN) datasets.Our new measure of PAC using Gamma GLMs and mutual information demonstrates a promising new way to compute PAC values using the full joint distribution on amplitude and phase. Our measure outperforms the most common existing measures of PAC, and show promising results in identifying time varying PAC in electrophysiological datasets. In addition, we provide for using our method with multiple comparisons and show that our measure potentially has more statistical power in electrophysiologic recordin","PeriodicalId":12363,"journal":{"name":"Frontiers in Computational Neuroscience","volume":null,"pages":null},"PeriodicalIF":3.2,"publicationDate":"2024-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141111319","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}