| Summary: | 碩士 === 中原大學 === 生物醫學工程研究所 === 104 === When sleep apnea occurs, it will increase heart rate, decrease pulse oximetry (SPO2) and cause sleep disturbances thereby affecting the quality of sleep. Diagnosis of sleep apnea is generally complicated. The purpose of this study is to develop an analysis software system based on heart rate variability to identify sleep apnea and to facilitate an early diagnosis of the symptom. We use the UCD Sleep Apnea Database to develop algorithms of identifying sleep apnea by analysis software. The function includes heart rate calculation, resampling, Fast Fourier Transform and power spectral density. The analysis software in this study has very high correlation with the Kubios analysis software (correlation coefficients >0.96). It shows highly reliability of analysis software. In part of algorithm, linear regression approach is used to determine sleep apnea syndrome. The linear regression correlation coefficient which is 0.4727 and the accuracy is 55%. In order to improve the accuracy, neural network training was carried out and it increased the accuracy of sleep apnea events to 71%. In conclusion, the present study developed an analysis software system based on heart rate variability to identify sleep apnea. The improvement of the analysis software algorithms will be continued to level up the accuracy and reliability of sleep apnea recognition.
|
|---|
