Neuroscience informatics最新文献_第5页

Reinforcement learning in artificial intelligence and neurobiology 人工智能和神经生物学中的强化学习

Neuroscience informatics

Pub Date : 2025-09-01 Epub Date: 2025-07-22 DOI: 10.1016/j.neuri.2025.100220

Tursun Alkam, Andrew H Van Benschoten, Ebrahim Tarshizi

Reinforcement learning (RL), a computational framework rooted in behavioral psychology, enables agents to learn optimal actions through trial and error. It now powers intelligent systems across domains such as autonomous driving, robotics, and logistics, solving tasks once thought to require human cognition. As RL reshapes artificial intelligence (AI), it raises a critical question in neuroscience: does the brain learn through similar mechanisms? Growing evidence suggests it does.

To bridge this interdisciplinary gap, this review introduces core RL concepts to neuroscientists and clinicians with limited AI exposure. We outline the agent–environment interaction loop and describe key architectures including model-free, model-based, and meta-RL. We then examine how advances in deep RL have generated testable hypotheses about neural computation and behavior. In parallel, we discuss how neurobiological findings, especially the role of dopamine in encoding reward prediction errors, have inspired biologically grounded RL models. Empirical studies reveal neural correlates of RL algorithms in the basal ganglia, prefrontal cortex, and hippocampus, supporting their roles in planning, memory, and decision-making. We also highlight clinical applications, including how RL frameworks are used to model cognitive decline and psychiatric disorders, while acknowledging limitations in scaling RL to biological complexity.

Looking ahead, RL offers powerful tools for understanding brain function, guiding brain–machine interfaces, and personalizing psychiatric treatment. The convergence of RL and neuroscience offers a promising interdisciplinary lens for advancing our understanding of learning and decision-making in both artificial agents and the human brain.

强化学习（RL）是一种植根于行为心理学的计算框架，它使代理能够通过试错来学习最佳行为。它现在为自动驾驶、机器人和物流等领域的智能系统提供动力，解决了曾经被认为需要人类认知的任务。随着强化学习重塑人工智能（AI），它提出了神经科学中的一个关键问题：大脑是否通过类似的机制进行学习？越来越多的证据表明确实如此。为了弥合这一跨学科的差距，本综述向人工智能接触有限的神经科学家和临床医生介绍了核心RL概念。我们概述了代理-环境交互循环，并描述了包括无模型、基于模型和元强化学习在内的关键体系结构。然后，我们研究了深度强化学习的进展如何产生关于神经计算和行为的可测试假设。同时，我们讨论了神经生物学的发现，特别是多巴胺在编码奖励预测错误中的作用，如何启发了基于生物学的RL模型。实证研究揭示了RL算法在基底神经节、前额叶皮层和海马体中的神经关联，支持它们在计划、记忆和决策中的作用。我们还强调了临床应用，包括RL框架如何用于模拟认知能力下降和精神疾病，同时承认将RL扩展到生物复杂性的局限性。展望未来，强化学习为理解大脑功能、指导脑机接口和个性化精神治疗提供了强大的工具。强化学习和神经科学的融合提供了一个很有前途的跨学科视角，可以促进我们对人工智能体和人脑中学习和决策的理解。

{"title":"Reinforcement learning in artificial intelligence and neurobiology","authors":"Tursun Alkam, Andrew H Van Benschoten, Ebrahim Tarshizi","doi":"10.1016/j.neuri.2025.100220","DOIUrl":"10.1016/j.neuri.2025.100220","url":null,"abstract":"<div><div>Reinforcement learning (RL), a computational framework rooted in behavioral psychology, enables agents to learn optimal actions through trial and error. It now powers intelligent systems across domains such as autonomous driving, robotics, and logistics, solving tasks once thought to require human cognition. As RL reshapes artificial intelligence (AI), it raises a critical question in neuroscience: does the brain learn through similar mechanisms? Growing evidence suggests it does.</div><div>To bridge this interdisciplinary gap, this review introduces core RL concepts to neuroscientists and clinicians with limited AI exposure. We outline the agent–environment interaction loop and describe key architectures including model-free, model-based, and meta-RL. We then examine how advances in deep RL have generated testable hypotheses about neural computation and behavior. In parallel, we discuss how neurobiological findings, especially the role of dopamine in encoding reward prediction errors, have inspired biologically grounded RL models. Empirical studies reveal neural correlates of RL algorithms in the basal ganglia, prefrontal cortex, and hippocampus, supporting their roles in planning, memory, and decision-making. We also highlight clinical applications, including how RL frameworks are used to model cognitive decline and psychiatric disorders, while acknowledging limitations in scaling RL to biological complexity.</div><div>Looking ahead, RL offers powerful tools for understanding brain function, guiding brain–machine interfaces, and personalizing psychiatric treatment. The convergence of RL and neuroscience offers a promising interdisciplinary lens for advancing our understanding of learning and decision-making in both artificial agents and the human brain.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 3","pages":"Article 100220"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144713172","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Predicting stroke with machine learning techniques in a sub-Saharan African population 用机器学习技术预测撒哈拉以南非洲人口中风

Neuroscience informatics

Pub Date : 2025-09-01 Epub Date: 2025-06-17 DOI: 10.1016/j.neuri.2025.100216

Benjamin Segun Aribisala , Deirdre Edward , Godwin Ogbole , Onoja M. Akpa , Segun Ayilara , Fred Sarfo , Olusola Olabanjo , Adekunle Fakunle , Babafemi Oluropo Macaulay , Joseph Yaria , Joshua Akinyemi , Albert Akpalu , Kolawole Wahab , Reginald Obiako , Morenikeji Komolafe , Lukman Owolabi , Godwin Osaigbovo , Akinkunmi Paul Okekunle , Arti Singh , Philip Ibinaye , Mayowa Owolabi

Background

Stroke is the second leading cause of death and the third leading cause of disability globally, including Africa, which bears its largest burden. Accurate models are needed in Africa to predict and prevent stroke occurrence. The aim of this study was to identify the best machine learning (ML) algorithm for stroke prediction.

Methods

We assessed medical data of 4,236 subjects comprising 2,118 stroke patients and 2,118 controls from the SIREN database. Sixteen established vascular risk factors were evaluated in this study. These are addition of salt to food at table during eating, cardiac disease, diabetes mellitus, dyslipidemia, education, family history of cardiovascular disease, hypertension, income, low green leafy vegetable consumption, obesity, physical inactivity, regular meat consumption, regular sugar consumption, smoking, stress and use of tobacco. From these, we also selected the 11 topmost risk factors using Population-Attributable Risk ranking. Eleven ML models were built and empirically investigated using the 16 and the 11 risk factors.

Results

Our results showed that the 16 features-based classification (maximum AUC of 82.32%) had a slightly better performance than the 11 feature-based (maximum AUC 81.17%) algorithm. The result also showed that Artificial Neural Network (ANN) had the best performance amongst eleven algorithms investigated with AUC of 82.32%, sensitivity of 71.23%, specificity of 80.00%.

Conclusion

Machine Learning algorithms predicted stroke occurrence employing major risk factors in Sub-Saharan Africa better than regression models. Machine Learning, especially Artificial Neural Network, is recommended to enhance Afrocentric stroke prediction models for stroke risk factor quantification and control in Africa.

中风是全球第二大致死原因和第三大致残原因，非洲也是其中之一，其负担最大。非洲需要准确的模型来预测和预防中风的发生。本研究的目的是确定用于中风预测的最佳机器学习（ML）算法。方法对来自SIREN数据库的2,118例脑卒中患者和2,118例对照组的4,236例受试者的医学资料进行评估。本研究评估了16个已确定的血管危险因素。这些因素包括：吃饭时在食物中添加盐、心脏病、糖尿病、血脂异常、教育程度、心血管疾病家族史、高血压、收入、绿叶蔬菜摄入量低、肥胖、缺乏体育锻炼、经常吃肉、经常吃糖、吸烟、压力和使用烟草。从这些因素中，我们还使用人口归因风险排名选择了11个最重要的风险因素。建立了11个ML模型，并对16个和11个危险因素进行了实证研究。结果基于16个特征的分类算法（最大AUC为82.32%）的分类性能略好于基于11个特征的分类算法（最大AUC为81.17%）。人工神经网络（Artificial Neural Network， ANN）的AUC为82.32%，灵敏度为71.23%，特异性为80.00%，在11种算法中表现最佳。结论机器学习算法预测撒哈拉以南非洲地区卒中发生的主要危险因素优于回归模型。建议使用机器学习，特别是人工神经网络来增强以非洲为中心的中风预测模型，用于非洲中风风险因素的量化和控制。

{"title":"Predicting stroke with machine learning techniques in a sub-Saharan African population","authors":"Benjamin Segun Aribisala , Deirdre Edward , Godwin Ogbole , Onoja M. Akpa , Segun Ayilara , Fred Sarfo , Olusola Olabanjo , Adekunle Fakunle , Babafemi Oluropo Macaulay , Joseph Yaria , Joshua Akinyemi , Albert Akpalu , Kolawole Wahab , Reginald Obiako , Morenikeji Komolafe , Lukman Owolabi , Godwin Osaigbovo , Akinkunmi Paul Okekunle , Arti Singh , Philip Ibinaye , Mayowa Owolabi","doi":"10.1016/j.neuri.2025.100216","DOIUrl":"10.1016/j.neuri.2025.100216","url":null,"abstract":"<div><h3>Background</h3><div>Stroke is the second leading cause of death and the third leading cause of disability globally, including Africa, which bears its largest burden. Accurate models are needed in Africa to predict and prevent stroke occurrence. The aim of this study was to identify the best machine learning (ML) algorithm for stroke prediction.</div></div><div><h3>Methods</h3><div>We assessed medical data of 4,236 subjects comprising 2,118 stroke patients and 2,118 controls from the SIREN database. Sixteen established vascular risk factors were evaluated in this study. These are addition of salt to food at table during eating, cardiac disease, diabetes mellitus, dyslipidemia, education, family history of cardiovascular disease, hypertension, income, low green leafy vegetable consumption, obesity, physical inactivity, regular meat consumption, regular sugar consumption, smoking, stress and use of tobacco. From these, we also selected the 11 topmost risk factors using Population-Attributable Risk ranking. Eleven ML models were built and empirically investigated using the 16 and the 11 risk factors.</div></div><div><h3>Results</h3><div>Our results showed that the 16 features-based classification (maximum AUC of 82.32%) had a slightly better performance than the 11 feature-based (maximum AUC 81.17%) algorithm. The result also showed that Artificial Neural Network (ANN) had the best performance amongst eleven algorithms investigated with AUC of 82.32%, sensitivity of 71.23%, specificity of 80.00%.</div></div><div><h3>Conclusion</h3><div>Machine Learning algorithms predicted stroke occurrence employing major risk factors in Sub-Saharan Africa better than regression models. Machine Learning, especially Artificial Neural Network, is recommended to enhance Afrocentric stroke prediction models for stroke risk factor quantification and control in Africa.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 3","pages":"Article 100216"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144322809","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Multimodal lightweight neural network for Alzheimer's disease diagnosis integrating neuroimaging and cognitive scores 综合神经影像学和认知评分的阿尔茨海默病多模态轻量级神经网络诊断

Neuroscience informatics

Pub Date : 2025-09-01 Epub Date: 2025-07-10 DOI: 10.1016/j.neuri.2025.100218

Bhoomi Gupta , Ganesh Kanna Jegannathan , Mohammad Shabbir Alam , Kottala Sri Yogi , Janjhyam Venkata Naga Ramesh , Vemula Jasmine Sowmya , Isa Bayhan

Conventional single-modal approaches for auxiliary diagnosis of Alzheimer's disease (AD) face several limitations, including insufficient availability of expertly annotated imaging datasets, unstable feature extraction, and high computational demands. To address these challenges, we propose Light-Mo-DAD, a lightweight multimodal diagnostic neural network designed to integrate MRI, PET imaging, and neuropsychological assessment scores for enhanced AD detection. In the neuroimaging feature extraction module, redundancy-reduced convolutional operations are employed to capture fine-grained local features, while a global filtering mechanism enables the extraction of holistic spatial patterns. Multimodal feature fusion is achieved through spatial image registration and summation, allowing for effective integration of structural and functional imaging modalities. The neurocognitive feature extraction module utilizes depthwise separable convolutions to process cognitive assessment data, which are then fused with multimodal imaging features. To further enhance the model's discriminative capacity, transfer learning techniques are applied. A multilayer perceptron (MLP) classifier is incorporated to capture complex feature interactions and improve diagnostic precision. Evaluation on the ADNI dataset demonstrates that Light-Mo-DAD achieves 98.0% accuracy, 98.5% sensitivity, and 97.5% specificity, highlighting its robustness in early AD detection. These results suggest that the proposed architecture not only enhances diagnostic accuracy but also offers strong potential for real-time, mobile deployment in clinical settings, supporting neurologists in efficient and reliable Alzheimer's diagnosis.

传统的用于阿尔茨海默病（AD）辅助诊断的单模态方法面临一些限制，包括专业注释的成像数据集可用性不足，特征提取不稳定以及计算需求高。为了解决这些挑战，我们提出了Light-Mo-DAD，这是一个轻量级的多模态诊断神经网络，旨在整合MRI， PET成像和神经心理学评估评分，以增强AD的检测。在神经成像特征提取模块中，采用了减少冗余的卷积运算来捕获细粒度的局部特征，同时采用了全局过滤机制来提取整体空间模式。通过空间图像配准和求和实现多模态特征融合，从而实现结构和功能成像模式的有效整合。神经认知特征提取模块利用深度可分离卷积来处理认知评估数据，然后将其与多模态成像特征融合。为了进一步提高模型的判别能力，本文采用了迁移学习技术。采用多层感知器（MLP）分类器捕获复杂的特征交互，提高诊断精度。对ADNI数据集的评估表明，Light-Mo-DAD的准确率为98.0%，灵敏度为98.5%，特异性为97.5%，突出了其在早期AD检测中的稳健性。这些结果表明，所提出的架构不仅提高了诊断的准确性，而且为临床环境中的实时、移动部署提供了强大的潜力，支持神经科医生高效、可靠地诊断阿尔茨海默病。

{"title":"Multimodal lightweight neural network for Alzheimer's disease diagnosis integrating neuroimaging and cognitive scores","authors":"Bhoomi Gupta , Ganesh Kanna Jegannathan , Mohammad Shabbir Alam , Kottala Sri Yogi , Janjhyam Venkata Naga Ramesh , Vemula Jasmine Sowmya , Isa Bayhan","doi":"10.1016/j.neuri.2025.100218","DOIUrl":"10.1016/j.neuri.2025.100218","url":null,"abstract":"<div><div>Conventional single-modal approaches for auxiliary diagnosis of Alzheimer's disease (AD) face several limitations, including insufficient availability of expertly annotated imaging datasets, unstable feature extraction, and high computational demands. To address these challenges, we propose Light-Mo-DAD, a lightweight multimodal diagnostic neural network designed to integrate MRI, PET imaging, and neuropsychological assessment scores for enhanced AD detection. In the neuroimaging feature extraction module, redundancy-reduced convolutional operations are employed to capture fine-grained local features, while a global filtering mechanism enables the extraction of holistic spatial patterns. Multimodal feature fusion is achieved through spatial image registration and summation, allowing for effective integration of structural and functional imaging modalities. The neurocognitive feature extraction module utilizes depthwise separable convolutions to process cognitive assessment data, which are then fused with multimodal imaging features. To further enhance the model's discriminative capacity, transfer learning techniques are applied. A multilayer perceptron (MLP) classifier is incorporated to capture complex feature interactions and improve diagnostic precision. Evaluation on the ADNI dataset demonstrates that Light-Mo-DAD achieves 98.0% accuracy, 98.5% sensitivity, and 97.5% specificity, highlighting its robustness in early AD detection. These results suggest that the proposed architecture not only enhances diagnostic accuracy but also offers strong potential for real-time, mobile deployment in clinical settings, supporting neurologists in efficient and reliable Alzheimer's diagnosis.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 3","pages":"Article 100218"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144634328","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

EEG–fNIRS signal integration for motor imagery classification using deep learning and evidence theory 基于深度学习和证据理论的EEG-fNIRS信号集成运动图像分类

Neuroscience informatics

Pub Date : 2025-09-01 Epub Date: 2025-06-18 DOI: 10.1016/j.neuri.2025.100214

Mohammed E. Seno , Niladri Maiti , Maulik Patel , Mihirkumar M. Patel , Kalpesh B. Chaudhary , Ashish Pasaya , Babacar Toure

To address the limitations of traditional unimodal brain-computer interface BCI) technologies based on electroencephalography (EEG) such as low spatial resolution and high susceptibility to noise an increasing number of neuroscience-driven studies have begun to focus on BCI systems that fuse EEG signals with functional near-infrared spectroscopy (fNIRS) signals. However, integrating these two heterogeneous neurophysiological signals presents significant challenges. In this work, we propose an innovative end-to-end signal fusion method based on deep learning and evidence theory for motor imagery (MI) classification within the neuroscience domain. For EEG signals, spatiotemporal features are extracted using dual-scale temporal convolution and depthwise separable convolution, and a hybrid attention module is introduced to enhance the network's sensitivity to salient neural patterns. For fNIRS signals, spatial convolution across all channels is employed to explore activation differences among brain regions, and parallel temporal convolution combined with a gated recurrent unit (GRU) captures richer temporal dynamics of the hemodynamic response. At the decision fusion stage, decision outputs from both modalities are first quantified using Dirichlet distribution parameter estimation to model uncertainty, followed by a two-layer reasoning process using Dempster-Shafer Theory (DST) to fuse evidence from basic belief assignment (BBA) methods and both modalities. Experimental evaluation on the publicly available TU-Berlin-A dataset demonstrates the effectiveness of the proposed model, achieving an average accuracy of 83.26%, representing a 3.78% improvement over state-of-the-art methods. These results provide new insights and methodologies for neuroscience-inspired multimodal BCI systems integrating EEG and fNIRS signals.

为了解决传统的基于脑电图（EEG）的单峰脑机接口（BCI）技术的局限性，如低空间分辨率和高噪声敏感性，越来越多的神经科学驱动的研究开始关注将EEG信号与功能近红外光谱（fNIRS）信号融合在一起的BCI系统。然而，整合这两种异质的神经生理信号提出了重大挑战。在这项工作中，我们提出了一种基于深度学习和证据理论的创新端到端信号融合方法，用于神经科学领域的运动图像（MI）分类。对于脑电信号，采用双尺度时间卷积和深度可分卷积提取时空特征，并引入混合注意模块增强网络对显著神经模式的敏感性。对于fNIRS信号，采用跨所有通道的空间卷积来探索脑区域之间的激活差异，并行时间卷积结合门控循环单元（GRU）捕获更丰富的血流动力学响应的时间动态。在决策融合阶段，首先使用Dirichlet分布参数估计对两种模式的决策输出进行量化以建模不确定性，然后使用Dempster-Shafer理论（DST）进行两层推理过程，以融合来自基本信念分配（BBA）方法和两种模式的证据。在公开可用的TU-Berlin-A数据集上的实验评估证明了所提出模型的有效性，平均准确率为83.26%，比最先进的方法提高了3.78%。这些结果为神经科学启发的多模态BCI系统集成EEG和fNIRS信号提供了新的见解和方法。

{"title":"EEG–fNIRS signal integration for motor imagery classification using deep learning and evidence theory","authors":"Mohammed E. Seno , Niladri Maiti , Maulik Patel , Mihirkumar M. Patel , Kalpesh B. Chaudhary , Ashish Pasaya , Babacar Toure","doi":"10.1016/j.neuri.2025.100214","DOIUrl":"10.1016/j.neuri.2025.100214","url":null,"abstract":"<div><div>To address the limitations of traditional unimodal brain-computer interface BCI) technologies based on electroencephalography (EEG) such as low spatial resolution and high susceptibility to noise an increasing number of neuroscience-driven studies have begun to focus on BCI systems that fuse EEG signals with functional near-infrared spectroscopy (fNIRS) signals. However, integrating these two heterogeneous neurophysiological signals presents significant challenges. In this work, we propose an innovative end-to-end signal fusion method based on deep learning and evidence theory for motor imagery (MI) classification within the neuroscience domain. For EEG signals, spatiotemporal features are extracted using dual-scale temporal convolution and depthwise separable convolution, and a hybrid attention module is introduced to enhance the network's sensitivity to salient neural patterns. For fNIRS signals, spatial convolution across all channels is employed to explore activation differences among brain regions, and parallel temporal convolution combined with a gated recurrent unit (GRU) captures richer temporal dynamics of the hemodynamic response. At the decision fusion stage, decision outputs from both modalities are first quantified using Dirichlet distribution parameter estimation to model uncertainty, followed by a two-layer reasoning process using Dempster-Shafer Theory (DST) to fuse evidence from basic belief assignment (BBA) methods and both modalities. Experimental evaluation on the publicly available TU-Berlin-A dataset demonstrates the effectiveness of the proposed model, achieving an average accuracy of 83.26%, representing a 3.78% improvement over state-of-the-art methods. These results provide new insights and methodologies for neuroscience-inspired multimodal BCI systems integrating EEG and fNIRS signals.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 3","pages":"Article 100214"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144470992","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Short-window EEG-based auditory attention decoding for neuroadaptive hearing support for smart healthcare 基于短窗口脑电图的听觉注意解码用于智能医疗的神经适应性听力支持

Neuroscience informatics

Pub Date : 2025-09-01 Epub Date: 2025-07-22 DOI: 10.1016/j.neuri.2025.100222

Ihtiram Raza Khan , Sheng-Lung Peng , Rupali Mahajan , Rajesh Dey

Background

Selective auditory attention the brain's ability to focus on a specific speaker in multi-talker environments is often compromised in individuals with auditory or neurological disorders. While Auditory Attention Decoding (AAD) using EEG has shown promise in detecting attentional focus, existing models primarily utilize temporal or spectral features, often neglecting the synergistic relationships across time, space, and frequency. This limitation significantly reduces decoding accuracy, particularly in short decision windows, which are crucial for real-time applications like neuro-steered hearing aids. This study is to enhance short-window AAD performance by fully leveraging multi-dimensional EEG characteristics.

Methods

To address this, we propose TSF-AADNet, a novel neural framework that integrates temporal–spatial and frequency–spatial features using dual-branch architectures and advanced attention-based fusion.

Results

Tested on KULeuven and DTU datasets, TSF-AADNet achieves 91.8% and 81.1% accuracy at 0.1-second windows—outperforming the state-of-the-art by up to 7.99%.

Conclusions

These results demonstrate the model's potential in enabling precise, real-time attention tracking for hearing impairment diagnostics and next-generation neuroadaptive auditory prosthetics.

选择性听觉注意在多说话的环境中，大脑专注于特定说话者的能力在听觉或神经障碍的个体中经常受到损害。虽然利用脑电图进行听觉注意解码（AAD）在检测注意焦点方面显示出前景，但现有模型主要利用时间或频谱特征，往往忽略了时间、空间和频率之间的协同关系。这种限制大大降低了解码的准确性，特别是在短决策窗口中，这对于神经导向助听器等实时应用至关重要。本研究旨在充分利用脑电图的多维特征，提高短窗口AAD的性能。为了解决这个问题，我们提出了一种新的神经框架TSF-AADNet，它使用双分支架构和先进的基于注意力的融合技术集成了时空和频率空间特征。结果在KULeuven和DTU数据集上测试，TSF-AADNet在0.1秒窗口下的准确率分别达到91.8%和81.1%，比目前最先进的准确率高出7.99%。这些结果证明了该模型在精确、实时的注意力跟踪听力障碍诊断和下一代神经适应性听觉假肢方面的潜力。

{"title":"Short-window EEG-based auditory attention decoding for neuroadaptive hearing support for smart healthcare","authors":"Ihtiram Raza Khan , Sheng-Lung Peng , Rupali Mahajan , Rajesh Dey","doi":"10.1016/j.neuri.2025.100222","DOIUrl":"10.1016/j.neuri.2025.100222","url":null,"abstract":"<div><h3>Background</h3><div>Selective auditory attention the brain's ability to focus on a specific speaker in multi-talker environments is often compromised in individuals with auditory or neurological disorders. While Auditory Attention Decoding (AAD) using EEG has shown promise in detecting attentional focus, existing models primarily utilize temporal or spectral features, often neglecting the synergistic relationships across time, space, and frequency. This limitation significantly reduces decoding accuracy, particularly in short decision windows, which are crucial for real-time applications like neuro-steered hearing aids. This study is to enhance short-window AAD performance by fully leveraging multi-dimensional EEG characteristics.</div></div><div><h3>Methods</h3><div>To address this, we propose TSF-AADNet, a novel neural framework that integrates temporal–spatial and frequency–spatial features using dual-branch architectures and advanced attention-based fusion.</div></div><div><h3>Results</h3><div>Tested on KULeuven and DTU datasets, TSF-AADNet achieves 91.8% and 81.1% accuracy at 0.1-second windows—outperforming the state-of-the-art by up to 7.99%.</div></div><div><h3>Conclusions</h3><div>These results demonstrate the model's potential in enabling precise, real-time attention tracking for hearing impairment diagnostics and next-generation neuroadaptive auditory prosthetics.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 3","pages":"Article 100222"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144696611","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Impact of physical activity and cardiorespiratory fitness on brain morphology among overweight and obese populations: A systematic review and meta-analysis of neuroimaging studies 身体活动和心肺健康对超重和肥胖人群脑形态的影响：神经影像学研究的系统回顾和荟萃分析

Neuroscience informatics

Pub Date : 2025-09-01 Epub Date: 2025-08-05 DOI: 10.1016/j.neuri.2025.100224

Dilip Shettigar , Suresh Sukumar , Rajagopal Kadavigere , K. Vaishali , Nitika C. Panakkal , Winniecia Dkhar , Abhimanyu Pradhan , Baskaran Chandrasekaran , Hari Prakash Palaniswamy , Poovitha Shruthi Paramashiva , Sneha Ravichandran , Sathya Sabina Muthu , Koustubh Kamath

引用次数: 0

Bayesian Inference General Procedures for A Single-subject Test study 单受试者检验研究的贝叶斯推断一般程序

Neuroscience informatics

Pub Date : 2025-06-01 Epub Date: 2025-03-12 DOI: 10.1016/j.neuri.2025.100195

Jie Li , Gary Green , Sarah J.A. Carr , Peng Liu , Jian Zhang

Abnormality detection in identifying a single-subject which deviates from the majority of a control group dataset is a fundamental problem. Typically, the control group is characterised using standard Normal statistics, and the detection of a single abnormal subject is in that context. However, in many situations, the control group cannot be described by Normal statistics, making standard statistical methods inappropriate. This paper presents a Bayesian Inference General Procedures for A Single-subject Test (BIGPAST) designed to mitigate the effects of skewness under the assumption that the dataset of the control group comes from the skewed Student t distribution. BIGPAST operates under the null hypothesis that the single-subject follows the same distribution as the control group. We assess BIGPAST's performance against other methods through simulation studies. The results demonstrate that BIGPAST is robust against deviations from normality and outperforms the existing approaches in accuracy, nearest to the nominal accuracy 0.95. BIGPAST can reduce model misspecification errors under the skewed Student t assumption by up to 12 times, as demonstrated in Section 3.3. We apply BIGPAST to a Magnetoencephalography (MEG) dataset consisting of an individual with mild traumatic brain injury and an age and gender-matched control group. For example, the previous method failed to detect abnormalities in 8 brain areas, whereas BIGPAST successfully identified them, demonstrating its effectiveness in detecting abnormalities in a single-subject.

识别偏离对照组数据集大部分的单个受试者的异常检测是一个基本问题。通常，使用标准的正常统计来描述对照组的特征，并且在此背景下检测单个异常受试者。然而，在许多情况下，对照组不能用正常统计来描述，使得标准统计方法不合适。本文提出了一个单受试者测试的贝叶斯推断通用程序（BIGPAST），旨在减轻偏性的影响，假设对照组的数据集来自偏斜的Student t分布。BIGPAST在单一受试者遵循与对照组相同分布的零假设下运行。我们通过模拟研究来评估BIGPAST与其他方法的性能。结果表明，BIGPAST对偏离正态性具有鲁棒性，并且在精度上优于现有方法，最接近名义精度0.95。如3.3节所示，在倾斜的Student t假设下，BIGPAST可以将模型误规范误差减少12倍。我们将BIGPAST应用于脑磁图（MEG）数据集，该数据集由轻度创伤性脑损伤个体和年龄和性别匹配的对照组组成。例如，之前的方法未能检测到8个大脑区域的异常，而BIGPAST成功地识别了它们，证明了它在检测单个受试者异常方面的有效性。

{"title":"Bayesian Inference General Procedures for A Single-subject Test study","authors":"Jie Li , Gary Green , Sarah J.A. Carr , Peng Liu , Jian Zhang","doi":"10.1016/j.neuri.2025.100195","DOIUrl":"10.1016/j.neuri.2025.100195","url":null,"abstract":"<div><div>Abnormality detection in identifying a single-subject which deviates from the majority of a control group dataset is a fundamental problem. Typically, the control group is characterised using standard Normal statistics, and the detection of a single abnormal subject is in that context. However, in many situations, the control group cannot be described by Normal statistics, making standard statistical methods inappropriate. This paper presents a Bayesian Inference General Procedures for A Single-subject Test (BIGPAST) designed to mitigate the effects of skewness under the assumption that the dataset of the control group comes from the skewed Student <em>t</em> distribution. BIGPAST operates under the null hypothesis that the single-subject follows the same distribution as the control group. We assess BIGPAST's performance against other methods through simulation studies. The results demonstrate that BIGPAST is robust against deviations from normality and outperforms the existing approaches in accuracy, nearest to the nominal accuracy 0.95. BIGPAST can reduce model misspecification errors under the skewed Student <em>t</em> assumption by up to 12 times, as demonstrated in Section <span><span>3.3</span></span>. We apply BIGPAST to a Magnetoencephalography (MEG) dataset consisting of an individual with mild traumatic brain injury and an age and gender-matched control group. For example, the previous method failed to detect abnormalities in 8 brain areas, whereas BIGPAST successfully identified them, demonstrating its effectiveness in detecting abnormalities in a single-subject.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 2","pages":"Article 100195"},"PeriodicalIF":0.0,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143643547","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Non-invasive brain stimulation-based sleep stage classification using transcranial infrared based electrocardiogram 基于无创脑刺激的经颅红外心电图睡眠阶段分类

Neuroscience informatics

Pub Date : 2025-06-01 Epub Date: 2025-03-18 DOI: 10.1016/j.neuri.2025.100197

Janjhyam Venkata Naga Ramesh , Aadam Quraishi , Yassine Aoudni , Mustafa Mudhafar , Divya Nimma , Monika Bansal

Non-invasive brain stimulation (NIBS) techniques, such as transcranial infrared (tNIR) stimulation, offer promising advancements in sleep monitoring and regulation. To enhance sleep stage classification without relying on traditional polysomnography (PSG) systems, we propose a novel approach integrating single-channel electrocardiogram (ECG) signals, heart rate variability (HRV) features, and tNIR stimulation. The maximal overlap discrete wavelet transform (MODWT) is applied for multi-resolution analysis of ECG signals, followed by peak information extraction. Based on the first-order deviation of peak positions, multi-dimensional HRV features are extracted. To identify HRV features strongly associated with different sleep stages, we introduce a feature selection method combining the ReliefF algorithm and Gini index. The selected features are then processed using the INFO-ABC Logit Boost method to establish correlations between HRV dynamics and sleep stages. Experimental results on publicly available datasets demonstrate that the proposed model achieves an overall accuracy of 83.67%, a precision of 82.59%, a Kappa coefficient of 77.94%, and an F1-score of 82.97%. Compared with conventional sleep staging methods, our approach enhances sleep quality assessment and facilitates real-time, non-invasive monitoring in home and mobile healthcare settings, leveraging the potential of tNIR-based NIBS for sleep modulation.

非侵入性脑刺激（NIBS）技术，如经颅红外（tNIR）刺激，在睡眠监测和调节方面提供了有希望的进步。为了在不依赖传统多导睡眠图（PSG）系统的情况下增强睡眠阶段分类，我们提出了一种整合单通道心电图（ECG）信号、心率变异性（HRV）特征和tNIR刺激的新方法。采用最大重叠离散小波变换（MODWT）对心电信号进行多分辨率分析，提取峰值信息。基于峰值位置的一阶偏差，提取了多维HRV特征。为了识别与不同睡眠阶段密切相关的HRV特征，我们引入了一种结合ReliefF算法和基尼指数的特征选择方法。然后使用INFO-ABC Logit Boost方法对选定的特征进行处理，以建立HRV动态与睡眠阶段之间的相关性。在公开数据集上的实验结果表明，该模型的总体准确率为83.67%，精密度为82.59%，Kappa系数为77.94%，f1分数为82.97%。与传统的睡眠分期方法相比，我们的方法增强了睡眠质量评估，促进了家庭和移动医疗环境中的实时、无创监测，充分利用了基于tnir的NIBS在睡眠调节方面的潜力。

{"title":"Non-invasive brain stimulation-based sleep stage classification using transcranial infrared based electrocardiogram","authors":"Janjhyam Venkata Naga Ramesh , Aadam Quraishi , Yassine Aoudni , Mustafa Mudhafar , Divya Nimma , Monika Bansal","doi":"10.1016/j.neuri.2025.100197","DOIUrl":"10.1016/j.neuri.2025.100197","url":null,"abstract":"<div><div>Non-invasive brain stimulation (NIBS) techniques, such as transcranial infrared (tNIR) stimulation, offer promising advancements in sleep monitoring and regulation. To enhance sleep stage classification without relying on traditional polysomnography (PSG) systems, we propose a novel approach integrating single-channel electrocardiogram (ECG) signals, heart rate variability (HRV) features, and tNIR stimulation. The maximal overlap discrete wavelet transform (MODWT) is applied for multi-resolution analysis of ECG signals, followed by peak information extraction. Based on the first-order deviation of peak positions, multi-dimensional HRV features are extracted. To identify HRV features strongly associated with different sleep stages, we introduce a feature selection method combining the ReliefF algorithm and Gini index. The selected features are then processed using the INFO-ABC Logit Boost method to establish correlations between HRV dynamics and sleep stages. Experimental results on publicly available datasets demonstrate that the proposed model achieves an overall accuracy of 83.67%, a precision of 82.59%, a Kappa coefficient of 77.94%, and an F1-score of 82.97%. Compared with conventional sleep staging methods, our approach enhances sleep quality assessment and facilitates real-time, non-invasive monitoring in home and mobile healthcare settings, leveraging the potential of tNIR-based NIBS for sleep modulation.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 2","pages":"Article 100197"},"PeriodicalIF":0.0,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143681918","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Integration of software-based cognitive approaches and brain-like computer machinery for efficient cognitive computing 基于软件的认知方法与类脑计算机机器的集成，实现高效的认知计算

Neuroscience informatics

Pub Date : 2025-06-01 Epub Date: 2025-03-13 DOI: 10.1016/j.neuri.2025.100194

Chitrakant Banchhor , Manoj Kumar Rawat , Rahul Joshi , Dharmesh Dhabliya , Omkaresh Kulkarni , Sandeep Dwarkanath Pande , Umesh Pawar

The widespread adoption of the Internet has transformed various industries, driving significant systemic reforms across different sectors. This transformation has enhanced the Internet's role in information dissemination, resource sharing, and global connectivity, allowing for more efficient distribution of knowledge and services. The development of the Internet model and its research bring significant benefits from the network, enabling people to use and learn from it. However, the traditional education model provides only limited knowledge, restricting growth and progress. Moreover, there is a vast world of knowledge yet to be explored. Nowadays, with the help of network tools, people can understand the dynamics of the whole world and accept the culture and knowledge of different regions without going out. Throughout the study of English legacy problems in various countries, efficient learning methods and high levels of English skills are the goals pursued, while the traditional English model can't meet the students' learning needs in a short time. The model construction of data mining algorithm based on large open network courses is a model for solving legacy problems adopted both domestically and internationally. According to the survey data of universities in various countries, the use of data mining algorithm can fundamentally meet the student's desire and demand for English knowledge. This research, integrates the mining algorithm into English research, which will essentially improve the English legacy problems.

互联网的广泛应用改变了各行各业，推动了不同领域的重大系统性改革。这一转变增强了互联网在信息传播、资源共享和全球互联互通方面的作用，使知识和服务的分配更加有效。互联网模式的发展及其研究为网络带来了巨大的利益，使人们能够使用网络并从中学习。然而，传统的教育模式只提供有限的知识，制约了成长和进步。此外，还有一个广阔的知识世界有待探索。如今，借助网络工具，人们足不出户就能了解整个世界的动态，接受不同地区的文化和知识。纵观各国对英语遗留问题的研究，高效的学习方法和高水平的英语技能是追求的目标，而传统的英语模式在短时间内无法满足学生的学习需求。基于大型开放网络课程的数据挖掘算法模型构建是国内外普遍采用的解决遗留问题的模型。根据各国大学的调查数据，数据挖掘算法的使用可以从根本上满足学生对英语知识的渴望和需求。本研究将挖掘算法整合到英语研究中，将从根本上改善英语遗留问题。

{"title":"Integration of software-based cognitive approaches and brain-like computer machinery for efficient cognitive computing","authors":"Chitrakant Banchhor , Manoj Kumar Rawat , Rahul Joshi , Dharmesh Dhabliya , Omkaresh Kulkarni , Sandeep Dwarkanath Pande , Umesh Pawar","doi":"10.1016/j.neuri.2025.100194","DOIUrl":"10.1016/j.neuri.2025.100194","url":null,"abstract":"<div><div>The widespread adoption of the Internet has transformed various industries, driving significant systemic reforms across different sectors. This transformation has enhanced the Internet's role in information dissemination, resource sharing, and global connectivity, allowing for more efficient distribution of knowledge and services. The development of the Internet model and its research bring significant benefits from the network, enabling people to use and learn from it. However, the traditional education model provides only limited knowledge, restricting growth and progress. Moreover, there is a vast world of knowledge yet to be explored. Nowadays, with the help of network tools, people can understand the dynamics of the whole world and accept the culture and knowledge of different regions without going out. Throughout the study of English legacy problems in various countries, efficient learning methods and high levels of English skills are the goals pursued, while the traditional English model can't meet the students' learning needs in a short time. The model construction of data mining algorithm based on large open network courses is a model for solving legacy problems adopted both domestically and internationally. According to the survey data of universities in various countries, the use of data mining algorithm can fundamentally meet the student's desire and demand for English knowledge. This research, integrates the mining algorithm into English research, which will essentially improve the English legacy problems.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 2","pages":"Article 100194"},"PeriodicalIF":0.0,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143643546","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Analyzing infant cry to detect birth asphyxia using a hybrid CNN and feature extraction approach 利用混合CNN和特征提取方法分析婴儿哭声以检测出生窒息

Neuroscience informatics

Pub Date : 2025-06-01 Epub Date: 2025-02-21 DOI: 10.1016/j.neuri.2025.100193

Samrat Kumar Dey , Khandaker Mohammad Mohi Uddin , Arpita Howlader , Md. Mahbubur Rahman , Hafiz Md. Hasan Babu , Nitish Biswas , Umme Raihan Siddiqi , Badhan Mazumder

Asphyxia, a critical respiratory condition, poses significant risks to newborns and can lead to catastrophic outcomes. Early detection of asphyxia is crucial for reducing infant mortality rates. Traditional medical diagnosis methods can be time-consuming, whereas early detection through artificial intelligence (AI) can expedite the process and improve survival rates. Despite the importance of early asphyxia detection, existing methods are often delayed and not always effective. This research addresses the need for a faster, more accurate approach to detecting infant asphyxia using machine learning (ML) and deep learning (DL) techniques. This study aims to develop a robust AI-driven system to detect asphyxia in newborns using ML and DL models, focusing on improving accuracy and efficiency over traditional diagnostic methods. This study explores feature extraction using Mel-Frequency Cepstral Coefficients (MFCCs), where the features are categorized into time and frequency domains. Data preprocessing techniques, such as noise removal, handling missing values, outliers, and label encoding, are applied to ensure clean data. To address class imbalance, the Random Oversampling (ROS) technique is employed. Hyperparameter optimization is performed using GridSearchCV for various machine-learning models. Deep learning models, including custom artificial neural networks (ANN1) and convolutional neural networks (CNN1, CNN2), are introduced with hidden layers for improved performance. The performance of different ML and DL models is evaluated, with Logistic Regression (LR) achieving an accuracy of 99.16% and a 0.008% error rate. In comparison, ANN1 outperforms other DL models with an accuracy of 98.20% and a 0.018% error rate. The results demonstrate that both ML and DL techniques can significantly enhance early asphyxia detection in newborns. The Logistic Regression model offers the highest accuracy in machine learning, while ANN1 performs optimally in deep learning, suggesting their potential for deployment in clinical settings to improve neonatal care.

窒息是一种严重的呼吸系统疾病，对新生儿构成重大风险，并可能导致灾难性后果。早期发现窒息对降低婴儿死亡率至关重要。传统的医疗诊断方法可能很耗时，而通过人工智能（AI）进行的早期检测可以加快过程并提高生存率。尽管早期窒息检测的重要性，现有的方法往往是延迟的，并不总是有效的。本研究解决了使用机器学习（ML）和深度学习（DL）技术更快，更准确地检测婴儿窒息的方法的需求。本研究旨在开发一个强大的人工智能驱动系统，使用ML和DL模型检测新生儿窒息，重点是提高传统诊断方法的准确性和效率。本研究探索了使用Mel-Frequency倒谱系数（MFCCs）的特征提取，其中特征被分类为时域和频域。数据预处理技术，如去噪、处理缺失值、异常值和标签编码，被用于确保干净的数据。为了解决类不平衡问题，采用了随机过采样（ROS）技术。使用GridSearchCV对各种机器学习模型进行超参数优化。深度学习模型，包括自定义人工神经网络（ANN1）和卷积神经网络（CNN1, CNN2），引入了隐藏层以提高性能。对不同ML和DL模型的性能进行了评估，其中逻辑回归（LR）的准确率为99.16%，错误率为0.008%。相比之下，ANN1的准确率为98.20%，错误率为0.018%，优于其他DL模型。结果表明，ML和DL技术都能显著提高新生儿早期窒息的检测。逻辑回归模型在机器学习中提供了最高的准确性，而ANN1在深度学习中表现最佳，这表明它们有潜力在临床环境中部署，以改善新生儿护理。

{"title":"Analyzing infant cry to detect birth asphyxia using a hybrid CNN and feature extraction approach","authors":"Samrat Kumar Dey , Khandaker Mohammad Mohi Uddin , Arpita Howlader , Md. Mahbubur Rahman , Hafiz Md. Hasan Babu , Nitish Biswas , Umme Raihan Siddiqi , Badhan Mazumder","doi":"10.1016/j.neuri.2025.100193","DOIUrl":"10.1016/j.neuri.2025.100193","url":null,"abstract":"<div><div>Asphyxia, a critical respiratory condition, poses significant risks to newborns and can lead to catastrophic outcomes. Early detection of asphyxia is crucial for reducing infant mortality rates. Traditional medical diagnosis methods can be time-consuming, whereas early detection through artificial intelligence (AI) can expedite the process and improve survival rates. Despite the importance of early asphyxia detection, existing methods are often delayed and not always effective. This research addresses the need for a faster, more accurate approach to detecting infant asphyxia using machine learning (ML) and deep learning (DL) techniques. This study aims to develop a robust AI-driven system to detect asphyxia in newborns using ML and DL models, focusing on improving accuracy and efficiency over traditional diagnostic methods. This study explores feature extraction using Mel-Frequency Cepstral Coefficients (MFCCs), where the features are categorized into time and frequency domains. Data preprocessing techniques, such as noise removal, handling missing values, outliers, and label encoding, are applied to ensure clean data. To address class imbalance, the Random Oversampling (ROS) technique is employed. Hyperparameter optimization is performed using GridSearchCV for various machine-learning models. Deep learning models, including custom artificial neural networks (ANN1) and convolutional neural networks (CNN1, CNN2), are introduced with hidden layers for improved performance. The performance of different ML and DL models is evaluated, with Logistic Regression (LR) achieving an accuracy of 99.16% and a 0.008% error rate. In comparison, ANN1 outperforms other DL models with an accuracy of 98.20% and a 0.018% error rate. The results demonstrate that both ML and DL techniques can significantly enhance early asphyxia detection in newborns. The Logistic Regression model offers the highest accuracy in machine learning, while ANN1 performs optimally in deep learning, suggesting their potential for deployment in clinical settings to improve neonatal care.</div></div>","PeriodicalId":74295,"journal":{"name":"Neuroscience informatics","volume":"5 2","pages":"Article 100193"},"PeriodicalIF":0.0,"publicationDate":"2025-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143478679","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0