基於前額腦波訊號之自動化睡眠分期判讀系統

Abstract

標準睡眠檢查使用的多通道生理訊號監測儀，需要在受測者身上黏貼大量電極，不僅事前黏貼及事後的清潔相當麻煩，繁亂的線路也會對睡眠品質造成不良影響，不論是對醫事人員或是受測者而言，都是一個痛苦的過程，也嚴重限制了能進行睡眠檢查的環境。而由於床位的不足及睡眠資料判讀方法的繁複，檢查前後往往需要漫長的等候，這樣的等待對於睡眠品質完全沒有幫助，也相當的沒有效率。在醫療資源有限的情況下，若能利用簡化的檢查預先將不需進行完整睡眠檢查的病患剔除，對醫事人員及病患都會是個福音。本研究提出了利用前額腦波進行自動化睡眠分期的方法，以較少的電極以及自動化的判讀機制，利用多個分類器組成階層式分類架構，在缺乏眼動訊號而難以判讀快速動眼期的情況下，與專家判讀間達到了70%的一致性，並且可分辨Wake、N1、N2、N3及REM等五個不同睡眠期。不僅簡化了電極的黏貼清潔步驟，提供患者於家中自行進行初步檢查的可能性，也為受測者節省下大量時間金錢上的花費，並且可以省去許多不必要的醫療資源浪費。而簡化後的系統，更是可以大幅放寬了睡眠檢查對醫療環境的依賴，提供了居家長期監測睡眠品質的可能性。

Polysomnography (PSG) is a common used method to diagnose sleep problems. However, a standard sleep study is a time consuming manual process. One shall wait for few weeks before proceeding a sleep study and wait another week for the report. Besides the waiting, the wires attached to a subject, the conductive gel applied to the electrodes and the unfamiliar environment would all disturb one’s sleep. In this thesis, a light weight sleep stage scoring system using forehead channels is introduced. Multiple classifiers were combined into a hierarchical classification system to perform automatic sleep stage classification using only Fp1 and Fp2 EEG signals. This system can separate REM sleep without EOG signals and achieve about 70% of agreement with expert’s scoring result. This system could provide a preliminary result and help deciding if a standard sleep study is required. This system also provides the possibility of long term sleep monitoring at home.