A RUSBoosted tree method for k-complex detection using tunable Q-factor wavelet transform and multi-domain feature...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• K-complex detection is crucial for sleep studies but traditionally relies on time-consuming manual analysis by clinicians.
• Existing automated methods often struggle with imbalanced datasets, hindering their effectiveness.
• Imbalanced data presents a significant challenge in accurate K-complex detection.

Purpose of the Study:

• To develop an efficient and accurate automated method for K-complex detection.
• To address the challenge of imbalanced datasets in K-complex detection.
• To improve the diagnostic tools for sleep disorders.

Main Methods:

• Utilized electroencephalogram (EEG) signals for K-complex detection.
• Employed tunable Q-factor wavelet transform (TQWT) for signal decomposition.
• Applied multi-domain feature extraction and selection coupled with a RUSBoosted tree model.
• Implemented a consistency-based filter for feature selection.

Main Results:

• The proposed method achieved high performance in K-complex detection, with recall, AUC, and F10-score reaching 92.41 ± 7.47%, 95.4 ± 4.32%, and 83.13 ± 8.59% respectively in Scenario 1.
• The RUSBoosted tree model demonstrated superior performance compared to LDA, logistic regression, and linear SVM, particularly in recall.
• The method proved effective even with imbalanced datasets, yielding consistent results across different scenarios.

Conclusions:

• The RUSBoosted tree model offers a promising solution for K-complex detection, especially with imbalanced data.
• This automated approach can serve as an effective tool for clinicians in diagnosing and treating sleep disorders.
• The developed method enhances the efficiency and accuracy of sleep analysis.