LLM辅助的标签噪声校正用于基于EEG的睡眠分期分类
LLM-Assisted Label Noise Correction for EEG-Based Sleep Stage Classification
Elmira Salari · Shruti Kshirsagar
中文摘要
基于脑电图(EEG)信号的自动睡眠分期分类是临床睡眠医学中的关键任务。深度学习模型在该领域已展现良好前景，但其性能从根本上受限于训练标签质量——这些标签由人类专家手动标注，存在评分者间不一致和系统性标注错误。提出一种LLM辅助的标签噪声过滤流程，仅在训练阶段运行，在重新训练分类器前识别并移除不可靠的睡眠分期标注。具体而言，先在原始数据集上训练TinySleepNet以提取每个epoch的SoftMax置信度分数，再将低置信度epoch连同其EEG频谱特征和周围序列上下文提交给LLaMA 3.1。LLM依据AASM睡眠评分指南评估每个epoch，并给出RELIABLE或SUSPICIOUS的判定。可疑epoch从训练集中移除，TinySleepNet在清洗后的数据上重新训练。在Sleep-EDF Expanded数据集上使用20折交叉验证评估，该方法将总体准确率从81.8%提升至84.6%，Macro F1-score从75.6%提升至77.4%，在REM检测上增益尤为显著。关键在于，由于LLM仅在训练阶段运行，推理时保留轻量级TinySleepNet架构（1.3M参数），使该流程适用于实时可穿戴睡眠监测应用。
关键要点
- 01问题：EEG睡眠分期标签由人工标注，存在评分者间不一致与系统性错误，制约模型性能。
- 02方法：先用TinySleepNet提取低置信度epoch，交由LLaMA 3.1结合频谱特征与上下文按AASM指南判定并过滤。
- 03结果：Sleep-EDF Expanded上20折交叉验证，准确率81.8%升至84.6%，Macro F1从75.6%升至77.4%。
- 04结果：REM检测增益尤为显著。
- 05局限/优势：LLM仅用于训练，推理保留1.3M参数TinySleepNet，适合实时可穿戴监测。
解读
尚无解读。
原始英文摘要
arXiv:2611.08311v1 Announce Type: new Abstract: Automatic sleep stage classification from electroencephalography (EEG) signals is a critical task in clinical sleep medicine. While deep learning models have shown promising results in this domain, their performance is fundamentally constrained by the quality of training labels, which human experts manually annotate and that are subject to inter-rater disagreement and systematic labeling errors. In this paper, we propose a novel LLM-assisted label-noise filtering pipeline that operates exclusively during training to identify and remove unreliable sleep-stage annotations before retraining the classifier. Specifically, TinySleepNet is first trained on the original dataset to extract per-epoch SoftMax confidence scores, then low-confidence epochs are submitted to LLaMA 3.1 along with their EEG spectral features and surrounding sequence context. The LLM evaluates each epoch against AASM sleep-scoring guidelines and assigns a verdict of RELIABLE or SUSPICIOUS. Suspicious epochs are removed from the training set, and TinySleepNet is retrained on the cleaned data. Evaluated on the Sleep-EDF Expanded dataset using 20-fold cross-validation, our method improves overall accuracy from 81.8% to 84.6% and Macro F1-score from 75.6% to 77.4%, with particularly notable gains in REM detection. Crucially, because the LLM operates only during training, the lightweight TinySleepNet architecture (1.3M parameters) is preserved at inference, making the proposed pipeline suitable for real-time, wearable sleep monitoring applications.