Hybrid deep neural network combined with Transformer for sleep disorders classification using biological signal

Show simple item record
dc.contributor.authorPham, Duc Thien
dc.date.accessioned2025-11-12T10:23:08Z
dc.date.available2025-11-12T10:23:08Z
dc.date.issued2025
dc.description.abstract-translatedSleep disorders, including sleep apnea, significantly impact overall health and quality of life. Traditional diagnostic methods, such as polysomnography (PSG), are resource-intensive, requiring specialized equipment and expert analysis. Deep learning (DL) models have emerged as a promising alternative for automated sleep stage classification and sleep apnea detection, offering improved efficiency and accessibility. Method: This study proposes novel hybrid deep learning models incorporating Transformer architectures for sleep stage classification using single-channel EEG signal and sleep apnea detection using single-lead ECG signal. Our approach leverages the strengths of each model component: CNN or 1D-ResNet-SE for extracting spatial and channel-wise features, Transformers for capturing long-range dependencies, and LSTM for modeling temporal sequences and long-term dependencies. Result: The experimental results on the ISRUC sleep dataset and the Physionet Apnea ECG dataset demonstrate that the proposed models outperform state-of-the-art methods. Specifically, the CNN-Transformer-LSTM achieved 82.4% accuracy on ISRUC-S3 and 80.37% on ISRUC-S1 for sleep stage classification using the EEG F3-A2 channel. For sleep apnea classification, it attained the highest accuracy of 91.6%, followed by the CNN-Transformer at 91.4% and the 1D-ResNet-SE-Transformer at 91.0%. Conclusion: The results suggest that these proposed models provide an efficient and accurate solution for automated sleep stage classification and sleep apnea detection. It can significantly alleviate human clinicians’ workload by automating sleep assessment and diagnosis, enabling faster, more consistent, and efficient evaluations.en
dc.description.sponsorshipThis work was supported by Grant No. SGS-2025-022 New Data Processing Methods in Current Areas of Computer Science.
dc.format55 s.cs
dc.identifier.urihttp://hdl.handle.net/11025/63851
dc.language.isoenen
dc.publisherUniversity of West Bohemiaen
dc.rights© University of West Bohemiaen
dc.subjectporuchy spánkucs
dc.subjecthluboké učenícs
dc.subjecttransformátorcs
dc.subject.translatedsleep disordersen
dc.subject.translateddeep learningen
dc.subject.translatedtransformeren
dc.titleHybrid deep neural network combined with Transformer for sleep disorders classification using biological signalen
dc.typereporten
dc.typezprávacs
local.files.count1*
local.files.size3945816*
local.has.filesyes*

Files

Original bundle
Showing 1 - 1 out of 1 results
No Thumbnail Available
Name:
TR_2025_02.pdf
Size:
3.76 MB
Format:
Adobe Portable Document Format
Download
License bundle
Showing 1 - 1 out of 1 results
No Thumbnail Available
Name:
license.txt
Size:
1.71 KB
Format:
Item-specific license agreed upon to submission
Description:
Download

Collections

Reports (KIV)