Amplifying Pathological Detection in EEG Signaling Pathways through Cross-Dataset Transfer Learning DOI Creative Commons
Mohammad Javad Darvishi Bayazi, Mohammad Sajjad Ghaemi, Timothée Lesort

et al.

arXiv (Cornell University), Journal Year: 2023, Volume and Issue: unknown

Published: Jan. 1, 2023

Pathology diagnosis based on EEG signals and decoding brain activity holds immense importance in understanding neurological disorders. With the advancement of artificial intelligence methods machine learning techniques, potential for accurate data-driven diagnoses effective treatments has grown significantly. However, applying algorithms to real-world datasets presents diverse challenges at multiple levels. The scarcity labelled data, especially low regime scenarios with limited availability real patient cohorts due high costs recruitment, underscores vital deployment scaling transfer techniques. In this study, we explore a pathology classification task highlight effectiveness data model cross-dataset knowledge transfer. As such, observe varying performance improvements through scaling, indicating need careful evaluation labelling. Additionally, identify possible negative emphasize significance some key components overcome distribution shifts spurious correlations achieve positive We see improvement target (NMT) by using from source dataset (TUAB) when amount was available. Our findings indicate small generic (e.g. ShallowNet) performs well single dataset, however, larger TCN) better dataset.

Language: Английский

Multimodal Large Language Models in Healthcare: Applications, Challenges, and Future Outlook (Preprint) DOI Creative Commons
Rawan AlSaad, Alaa Abd‐Alrazaq, Sabri Boughorbel

et al.

Journal of Medical Internet Research, Journal Year: 2024, Volume and Issue: 26, P. e59505 - e59505

Published: Aug. 20, 2024

In the complex and multidimensional field of medicine, multimodal data are prevalent crucial for informed clinical decisions. Multimodal span a broad spectrum types, including medical images (eg, MRI CT scans), time-series sensor from wearable devices electronic health records), audio recordings heart respiratory sounds patient interviews), text notes research articles), videos surgical procedures), omics genomics proteomics). While advancements in large language models (LLMs) have enabled new applications knowledge retrieval processing field, most LLMs remain limited to unimodal data, typically text-based content, often overlook importance integrating diverse modalities encountered practice. This paper aims present detailed, practical, solution-oriented perspective on use (M-LLMs) field. Our investigation spanned M-LLM foundational principles, current potential applications, technical ethical challenges, future directions. By connecting these elements, we aimed provide comprehensive framework that links aspects M-LLMs, offering unified vision their care. approach guide both practical implementations M-LLMs care, positioning them as paradigm shift toward integrated, data–driven We anticipate this work will spark further discussion inspire development innovative approaches next generation systems.

Language: Английский

Citations

29
Role of deep learning in cognitive healthcare: wearable signal analysis, algorithms, benefits, and challenges DOI Creative Commons
Md. Sakib Bin Alam, Aiman Lameesa, Sadia Sharmin

et al.

Digital Communications and Networks, Journal Year: 2025, Volume and Issue: unknown

Published: April 1, 2025

Language: Английский

Citations

0
Automatic diagnostics of electroencephalography pathology based on multi-domain feature fusion DOI Creative Commons
Shimiao Chen, Dong Huang, Xinyue Liu

et al.

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(5), P. e0310348 - e0310348

Published: May 5, 2025

Electroencephalography (EEG) serves as a practical auxiliary tool deployed to diagnose diverse brain-related disorders owing its exceptional temporal resolution, non-invasive characteristics, and cost-effectiveness. In recent years, with the advancement of machine learning, automated EEG pathology diagnostics methods have flourished. However, most existing usually neglect crucial spatial correlations in multi-channel signals potential complementary information among different domain features, both which are keys improving discrimination performance. addition, latent redundant irrelevant features may cause overfitting, increased model complexity, other issues. response, we propose novel feature-based framework designed improve diagnostic accuracy pathology. This first applies multi-resolution decomposition technique statistical feature extractor construct salient time-frequency space. Then, distribution is channel-wise extracted from this space fuse thereby leveraging their complementarity fullest extent. Furthermore, eliminate redundancy irrelevancy, two-step dimension reduction strategy, including lightweight multi-view aggregation non-parametric significance analysis, devised pick out stronger discriminative ability. Comprehensive examinations Temple University Hospital Abnormal Corpus V. 2.0.0 demonstrate that our proposal outperforms state-of-the-art methods, highlighting significant clinically abnormality detection.

Language: Английский

Citations

0
Amplifying pathological detection in EEG signaling pathways through cross-dataset transfer learning DOI Creative Commons
Mohammad Javad Darvishi Bayazi,

Mohammad Sajjad Ghaemi,

Timothée Lesort

et al.

Computers in Biology and Medicine, Journal Year: 2023, Volume and Issue: 169, P. 107893 - 107893

Published: Dec. 30, 2023

Pathology diagnosis based on EEG signals and decoding brain activity holds immense importance in understanding neurological disorders. With the advancement of artificial intelligence methods machine learning techniques, potential for accurate data-driven diagnoses effective treatments has grown significantly. However, applying algorithms to real-world datasets presents diverse challenges at multiple levels. The scarcity labeled data, especially low regime scenarios with limited availability real patient cohorts due high costs recruitment, underscores vital deployment scaling transfer techniques. In this study, we explore a pathology classification task highlight effectiveness data model cross-dataset knowledge transfer. As such, observe varying performance improvements through scaling, indicating need careful evaluation labeling. Additionally, identify possible negative emphasize significance some key components overcome distribution shifts spurious correlations achieve positive We see improvement target (NMT) by using from source dataset (TUAB) when amount was available. Our findings demonstrated that small generic (e.g. ShallowNet) performs well single dataset, however, larger TCN) better leveraging more dataset.

Language: Английский

Citations

9
Pediatric and Adolescent Seizure Detection: A Machine Learning Approach Exploring the Influence of Age and Sex in Electroencephalogram Analysis DOI Creative Commons
Lan Wei, Catherine Mooney

BioMedInformatics, Journal Year: 2024, Volume and Issue: 4(1), P. 796 - 810

Published: March 6, 2024

Background: Epilepsy, a prevalent neurological disorder characterized by recurrent seizures affecting an estimated 70 million people worldwide, poses significant diagnostic challenge. EEG serves as important tool in identifying these seizures, but the manual examination of EEGs experts is time-consuming. To expedite this process, automated seizure detection methods have emerged powerful aids for expert analysis. It worth noting that while such are well-established adult EEGs, they been underdeveloped pediatric and adolescent EEGs. This study sought to address gap devising automatic system tailored data. Methods: Leveraging publicly available datasets, TUH CHB-MIT machine learning-based models were constructed. The dataset was divided into training (n = 118), validation 19), testing 37) subsets, with special attention ensure clear demarcation between individuals test sets preserve set’s independence. used external set. Age sex incorporated features investigate their potential influence on detection. Results: By leveraging 20 extracted from both time frequency domains, along age additional feature, method achieved accuracy 98.95% set 64.82% Our investigation revealed crucial factor accurate Conclusion: outcomes hold substantial promise supporting researchers clinicians engaged analysis

Language: Английский

Citations

2
ARNN: Attentive recurrent neural network for multi-channel EEG signals to identify epileptic seizures DOI
Salim Rukhsar, Anil Kumar Tiwari

Neurocomputing, Journal Year: 2024, Volume and Issue: unknown, P. 129203 - 129203

Published: Dec. 1, 2024

Language: Английский

Citations

1
Analysis and Comparison of Artifact Removal Techniques for Epilepsy EEG Signal DOI Open Access

G Mamatha,

S. A. Hariprasad

Indian Journal of Science and Technology, Journal Year: 2024, Volume and Issue: 17(23), P. 2370 - 2380

Published: May 28, 2024

Objective: Accurate epilepsy diagnosis demands precise EEG analysis, hindered by non-neuronal artifacts. This study evaluates artifact removal methods, specifically Independent Component Analysis (ICA) and Empirical Mode Decomposition (EMD), aiming to enhance signal quality. We introduce a hybrid approach, combining ICA EMD. Methods: EMD are applied preprocess recordings. Quantitative evaluation metrics, including Signal-to-Noise Ratio (SNR), Peak (PSNR), Mean Squared Error (MSE), Root (RMSE), Standard Deviation (SD), calculated compared for both methods. Findings: outperforms EMD, showing higher SNR PSNR, notably in BONN CHB-MIT datasets. achieves significant reductions MSE, RMSE, SD. The approach surpasses existing supported quantitative data. Novelty: Rigorous application of diverse datasets quantitatively establishes ICA's superiority. backed evidence, proves effective beyond EEG. Conclusion: abstract provides clear, support superiority the approach's efficacy, offering valuable insights into analysis. Keywords: Epilepsy, Artifact removal, EEG, ICA, DWT, Performance metrics

Language: Английский

Citations

0
Integrating CNN with EEG Signals for Emergency Distance Prediction DOI

S. Sivasaravana Babu,

T. R. Dinesh Kumar,

G. Saravana Kumar

et al.

Published: April 17, 2024

EEG signals are a useful tool for determining person's cognitive and attentional states since they show the electrical activity of brain. By analyzing using convolutional neural networks (CNN), it is possible to extract features that can be correlated with perceived distance from an emergency event. This paper proposes integrating CNN prediction. It involves several key steps. First, dataset Electroencephalography (EEG) individuals at varying distances event needs collected. These typically recorded specialized sensors placed on scalp. The procedure then advances decomposition db6 wavelets. input signal first broken down into frequency sub bands DB6 wavelet decomposition. Both high-frequency low-frequency components may captured during decomposing, allowing more thorough examination. To further break intrinsic mode functions (IMFs), empirical (EMD) applied. technique adaptively decomposes oscillatory distinct time scales. Energy power calculation methods applied quantify energy distribution across different components. Next, in order provide final outputs, models built trained understand correlation between associated finally executed Matlab program. CNN's accuracy 92.6%, while its specificity comparison 91%.

Language: Английский

Citations

0
Double Discrete Variational Autoencoder for Epileptic EEG Signals Classification DOI Creative Commons

Shufeng Liang,

Xin Zhang, Hulin Zhao

et al.

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 106567 - 106578

Published: Jan. 1, 2024

Electroencephalography (EEG) plays a key role in the clinical evaluation of epilepsy and provides strong support for treatment decisions. However, analyzing decoding EEG recordings is burdensome task neurologists experts. Existing automated detection techniques make considerable efforts feature engineering, but often fall short when it comes to representing complex patterns signals. Deep learning methods allow higher-order representations intricate pattern learning, experiencing explosive success performance diagnostics. In this paper, we propose novel Double Discrete Variational AutoEncoder (D2-VAE) network learn latent signals perform deep discretization. Specifically, method builds learnable codebook based on Vector Quantized (VQ-VAE) obtain generic representation signal. The discretization local signal obtained by queries, whereas global information characterized building histogram quantization patterns. Such local-global portrayal more attuned single-mode repetition multi-mode mixing that characterizes epileptic Multiple diagnostic tasks multiple metrics are used validate effectiveness proposed classification experimental results show D2-VAE possesses low-dimensional yet powerful quantitative representation, with significant improvement over existing methods.

Language: Английский

Citations

0
Automatic diagnostics of electroencephalography pathology based on multi-domain feature fusion DOI Creative Commons
Shimiao Chen, Dong Huang, Xinyue Liu

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 3, 2024

Abstract Electroencephalography (EEG) serves as a practical auxiliary tool deployed to diagnose diverse brain-related disorders owing its exceptional temporal resolution, non-invasive characteristics, and cost-effectiveness. In recent years, with the advancement of machine learning, automated EEG pathology diagnostics methods have flourished. However, most existing usually neglect crucial spatial correlations in multi-channel signals potential complementary information among different domain features, both which are keys improving discrimination performance. addition, latent redundant irrelevant features may cause overfitting, increased model complexity, other issues. response, we propose novel feature-based framework designed improve diagnostic accuracy pathology. This first applies multi-resolution decomposition technique statistical feature extractor construct salient time-frequency space. Then, distribution is channel-wise extracted from this space fuse thereby leveraging their complementarity fullest extent. Furthermore, eliminate redundancy irrelevancy, two-step dimension reduction strategy, including lightweight multi-view aggregation non-parametric significance analysis, devised pick out stronger discriminative ability. Comprehensive examinations Temple University Hospital Abnormal Corpus V. 2.0.0 demonstrate that our proposal outperforms state-of-the-art methods, highlighting significant clinically abnormality detection.

Language: Английский

Citations

0