Summary: | 碩士 === 國立成功大學 === 醫學資訊研究所 === 98 === It is found from the research that some disease and behavioural have relation with sleep quality and quantity, but some pathogenesis is not clear and lack of an effective clinical treatment, so requires an ideal animal model for study. Rats are often used as models because they are readily available and display electrical activity during sleep that has similarities with human sleep. Both rat and human, sleep state is divided into two categories: non-rapid eye movement (NREM) sleep and rapid eye movement (REM) sleep. In human, NREM from light to deep is subdivided as s1, s2 and SWS. In rats, according to Neckelmann et al. proposed rules in 1994, NREM is subdivided into light sleep , deep sleep and transition sleep.Visual scoring is time and energy consuming. Therefore, the aim of thesis is to propose a high accuracy and reliable automatic sleep staging method for rats.
The study is designed to an automated sleep scoring method by using features of the electroencephalogram (EEG) and on electrotromyogram (EMG) activity. For decrease the difference of subjects on the feature, we do normalization. Each 10-sec epoch, we calculate state indices to distinguish waking, NREM and REM state roughly first, and then according to five 2-sec epoch of index value and band power ratio scoring per 10-sec epoch to subdivide into waking, NREM 1, NREM 2 ,transition sleep and REM.
The system is validated with 16 recordings of 24-hour each by comparing with visual scoring for two raters. For scored stage are waking, NREM 1, NREM 2, transition sleep and REM, agreement between scorer consensus and automatic scoring is 92.5% (kappa = 0.88). Global agreement between scorer consensus and automatic scoring is 95.3% (kappa = 0.91) on waking/NREM/REM. This indicated that the performance is high accuracy and reliability.
|