Research on EEG Signal Processing and Pattern Recognition in Sleep Related Applications
Bibliographic record
Abstract
Sleep is one of the most important physiological process for human beings. Nowadays, however, an increasing number of people are suffering from sleep-related diseases and disorders. In order to diagnose sleep-related disorders, sleep is usually monitored using the polysomnography (PSG) devices which record electroencephalogram (EEG), electrocardiogram (ECG), electrooculogram (EOG), electromyogram (EMG), etc. \nAmong all these physiological signals, EEG is regarded as the gold standard signal for sleep-related analysis, which scores the electrical activity of the brain. EEG is aperiodic and non-stationary without distinct morphological features, and may vary considerably in amplitude and fluctuation across subjects, making the processing and analysis of EEG signal a challenging task. \nThe correlation between sleep and the rich biomedical information embedded in EEG has been explored deeply in current sleep-related analysis by means of complicated EEG signal processing and pattern recognition technologies. In order to automate clinical sleep-related diagnosis and provide a reliable analysis result based on EEG signals, this dissertation focuses on the research of EEG processing and pattern recognition technologies in the following three aspects that attract much attention in the field of sleep research: sleep spindle detection, sleep stage classification and sleep apnea detection. \nSleep spindle activities are usually stable for a subject yet may vary considerably across subjects. In order to boost generalization ability and robustness of the automatic sleep spindle detection method that screens out sleep spindle activities from raw central-channel EEG signals, an adaptive Teager energy parameters method is proposed. The method achieves a high recognition rate for spindle activities. \nSleep stage classification from EEG is the basis of sleep monitoring and evaluation, which segments a whole-night sleep into 30-second epochs and annotate each of them with one of five different sleep stages. We propose two different methods for different input modalities. For inputting single-channel EEG, a multimodal decomposition with a hidden Markov model (HMM) based refinement method is presented, which achieves an overall accuracy as high as 0.894 for Sleep-EDF database and 0.793 for MASS database (Montreal Archive of Sleep Studies). For inputting multi-channel EEG, a novel sleep stage classification method is proposed based on the features extracted from the covariance matrices mapping on Riemannian manifold using multi-channel EEG. An overall accuracy of 0.801 and a Cohen’s Kappa coefficient of 0.704 are obtained on MASS database. \nSleep apnea detection from EEG aims to identifying the seizure of sleep apnea from EEG signals. We propose a novel framework for this task. The single-channel EEG is firstly transformed into time-frequency images, and classified by a multi-scale parallel convolutional neural network (CNN). On MIT-BIH polysomnographic dataset, this method yields an overall accuracy of 0.891 for sleep apnea recognition. \nThis dissertation concludes our research on EEG signal processing and pattern recognition technologies, and explores the optimal solutions in aforementioned aspects: sleep spindle detection, sleep stage classification and sleep apnea detection. Research in this dissertation is expected to assist and improve the diagnosis and analysis of sleep-related pathologies.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.003 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".