- 學位論文
以多項式曲線擬合改善心房顫動病患之血壓量測準確度
Improve the Accuracy of Blood Pressure Measurement of Atrial fibrillation patients using polynomial curve fitting
摘要
心房顫動(atrial fibrillation)為常見的心律不整疾病之一。根據研究顯示，近50%的心房顫動病人患有高血壓。且心房顫動的發病率隨著年齡而增加，特別是65歲以後的族群。對心房顫動患者來說，有效的追蹤控制血壓是相當重要的。但心房顫動的發生，不規則的心律和心輸出量的改變，使得傳統的聽診法和示波震盪法在血壓量測上容易造成高、低估的情形，導致血壓量測不準確。 本研究針對具有心房顫動現象的族群，在進行導管電氣燒灼術的同時，同步量測侵入式血壓訊號和非侵入式的血壓震盪波形。分別紀錄心房顫動發生時和經過電燒手術後恢復成為竇性心律時這兩個狀態的血壓訊號。將量測到的血壓波形數據透過電腦進行離線訊號處理和分析。將非侵入式的血壓振盪波形封包進行多項式曲線擬合，以侵入式的血壓當作參考標準，尋找最合適的特徵比例計算血壓值。最後，將演算法實現在硬體裝置上，以提高心房顫動血壓量測的準確度。 本研究蒐集了36位受測者，平均年齡為54.53歲。結果發現心房顫動時取時間間隔為二十秒的平均數值可代表一分鐘平均數值。所以本研究以二十秒的平均和非侵入式血壓做比較，取最小平均誤差來決定特徵比例。在竇性心律時，收縮壓和舒張壓的特徵比例分別為0.32和0.86；而在心房顫動時，收縮壓和舒張壓的特徵比例分別為0.56和0.96。 最後比較系統裝置和Matlab、市售血壓計(WatchBP office AFIB, Microlife)、侵入式血壓量測的結果。在SR和AF狀態下，本系統量測的血壓數值(SBP、DBP、MBP)和Matlab計算的血壓數值皆沒有顯著上差異(P>0.05)，Matlab計算的血壓數值和侵入式血壓也均無顯著上的差異。 系統分類結果部份，以T index(0.895)當作判別AF和SR的參考依據，此裝置辨識心房顫動的靈敏性為100%、特異性為24%、準確性為59%。
並列摘要
Atrial fibrillation (AF) is the most common sustained arrhythmia. According to previous studies, approximately 50% of the patients with AF are hypertensives and patients who are older than 65 years the incidence of AF is associated with increased age. Effective follow-up and blood pressure (BP) control is important for patients with AF. The irregularity of the pulse rhythm and variable stroke volume during AF make conventional auscultatory and oscillometric BP measurement inaccurate. In this study, we measured invasive and non-invasive oscillometric BP signals, simultaneously, while patients were conducting electrical ablation. The BP waveforms during AF and sinus rhythm (SR) were recorded. These BP data were processed and analyzed off-line in a personal computer. Additionally, using the invasive BP as reference, polynomial curve fitting were implemented to determine the most suitable characteristic ratio for non-invasive oscillometric BP measurement. Finally, the algorithm is realized in hardware devices, in order to improve the BP measurements accuracy in AF. Thirty six subjects were recruited in this study, and the mean age is 54.53 years. It was found that the average BPs in 20s intervals are compatible to one minute average. Thus, 20s averages were used to compare with invasive BP measurement to establish the characteristic ratios for non-invasive BP measurement using the minimum mean error. During SR, the characteristic ratios for systolic and diastolic blood pressure are 0.32 and 0.86, respectively. On the other hand, during AF, the characteristic ratios for systolic and diastolic blood pressure are 0.56 and 0.96, respectively. Lastly, we compared the BP readings of proposed system, Matlab, electric sphygmomanometer (WatchBP office AFIB, Microlife), and invasive measurements. During AF and SR, the t-test results indicate that there was no significant different in systolic, diastolic, mean BPs obtain from the proposed system and Matlab (P>0.05); and no significant difference between Matlab and invasive were founded (P>0.05). In the results of classification, using T index (0.895) to partition AF and SR, the proposed system is able to identify AF with a sensitivity of 100%, specificity of 24%, and accuracy of 59%.