無袖帶連續血壓監測器之驗證方法：以A公司Accurate 24產品為例

Abstract

無袖帶血壓計( Cuffless Blood Pressure Monitor)，也被稱為無壓脈帶血壓量測裝置或無袖帶血壓監測器，是近年來出現的一種新技術，相較於傳統有袖帶壓脈的血壓量測方式，以這個新技術設計出來的產品能夠長時間內進行連續血壓監測，在早期診斷出潛在的高血壓患者，特別適用於需要長時間監控血壓的患者。 儘管無袖帶血壓計提供了方便的監測方法，但是能否如傳統的血壓計那麼準確，則需要再進行驗證。 目前業界缺乏公認可靠的驗證設備及方法，我們在實務上要驗證一個無袖帶血壓計的新產品時，便需要從頭開始規劃實驗方式，設計實驗輔具，並研究如何找出修正參數。 本論文將這個在實驗室裡採用何種理論依據，實驗過程，與實驗結論做詳實紀錄。

研究方法: 這個實驗設計的起始點，聚焦於脈搏波速度 (pulse wave velocity PWV) 的測量準確度，因為脈搏波速度的量測是影響無袖帶血壓計對於血壓估計準確度的重要因素。 為了取得脈搏波速度量測的數據，實驗開始時便需要先開發一個能夠模擬脈搏波的模擬器。它包含模擬人體心臟脈動的機電系統，以及模擬手臂與動脈的假體，兩個部分組合成具有血流動力學特性的局部脈搏波模擬器，來模擬人體經過心臟跳動產生的脈搏波傳遞到手腕肱動脈的狀態，而肱動脈便是此次待測的無袖帶血壓計要量測的位置。 以這個測量局部脈搏波速度 (local PWV) 的無袖帶血壓計作為測試儀器，來測量脈搏波模擬器的脈搏波速度(PWV)。 再以血流動力學模型來擬合無袖帶血壓計與脈搏波模擬器的結果，該模型可以快速校準無袖帶血壓計的血流動力學測量性能。

實驗結果: 在根據血流動力學取得上述量測的結果後，再使用多元線性回歸 (multiple linear regression MLR) 計算出無袖帶血壓計的校準模型。 分別實驗了使用多元線性回歸MLR模型校準前與校準後的對比，紀錄兩者所測得的脈搏波速度差異，其中無袖帶血壓計直接測量脈搏波模擬器透過血流動力學模型得出的脈搏波速度平均絕對誤差 (mean absolute error MAE)為0.77m/s，使用多元線性回歸模型校準後再進行測量平均絕對誤差減少至0.06m/s。 這表示根據血流動力學模型計算的血壓誤差，在血壓100mmHg到180mmHg之間無袖帶血壓計的測量誤差，在校準前的誤差會達到1.7mmHg到5.99mmHg，而校準後的測量誤差只有0.14mmHg到0.48mmHg，大幅提升量測準確度。

研究結論: 這項研究主要是圍繞在如何準確的測量出脈搏波開始的，這對於無袖帶血壓計的血壓量測準確度攸關重要。 為此設計了一款以血流動力學特性的脈搏波模擬器，同時找出標準的性能驗證方法。 將待測的無袖帶血壓計與脈搏波模擬器進行MLR多元線性回歸建模，接著以此脈搏波模擬器，測量出相同設計原理的無袖帶血壓計的特性和差異。 依本研究原理所設計的脈搏波模擬器適合用於大量生產，並能快速驗證出無袖帶血壓計的特性，因此也適用於工業生產時，做為產品檢驗的環節之一，能夠協助無袖帶血壓計的生產驗證與普及。

The Cuffless Blood Pressure Monitor, alternately referred to as the sleeveless blood pressure monitor or simply, the cuffless BP monitor, is an innovative technology that has emerged in recent years. Contrasted with traditional cuffed blood pressure measurement methods, devices designed using this new technology enable prolonged, continuous blood pressure monitoring. This proves pivotal in the early diagnosis of potential hypertensive patients and is especially suited for those requiring long-term monitoring. However, questions arise about their accuracy compared to the traditional cuffed devices. Our study sets out to rigorously validate this, particularly focusing on the pivotal role of pulse wave velocity (PWV) in determining the measurement accuracy of these cuffless devices.

Introduction: While cuffless blood pressure monitors offer a convenient means of monitoring, the question arises as to whether they can achieve the same level of accuracy as traditional cuffed blood pressure devices. Rigorous calibration validation is necessary to confirm this. Given the current industry's lack of universally accepted reliable equipment and validation methods, we have initiated a study to investigate the validation process for cuffless blood pressure monitors. This involves planning simulated experiments, designing the experimental procedures, identifying calibration parameters, and progressively documenting the research references and processes. In this paper document above study and present a further clinical study results using the cuffless blood pressure monitor calibrated by the equipment and method in the study.

Methodology: Our investigation began with a focus on the accuracy of measuring the pulse wave velocity (PWV). This metric is instrumental in influencing the precision of the cuffless BP device's blood pressure estimations. To obtain accurate PWV data, we devised a simulator capable of mimicking human pulse waves. This simulator comprises a mechatronic system emulating human heart pulses, combined with a prosthetic replicating the arm and arteries. Together, they simulate the state of a human pulse wave as it travels from the heart to the arm's arteries. Using the simulator's characteristics, we tested its PWV using a cuffless device, calibrating it with a hemodynamic model. This model rapidly adjusted the device's hemodynamic measurement capabilities.

Results: A multiple linear regression (MLR) model was first utilized to calibrate the cuffless BP monitor. A comparison of the PWV measurements taken directly from the pulse wave simulator and the hemodynamic model showed a mean absolute error (MAE) of 0.77 m/s. With MLR calibration, this error reduced to 0.06 m/s. This signifies that the blood pressure measurement error, based on the hemodynamic model and ranging between 100mmHg to 180mmHg, dropped from a pre-calibration range of 1.7mmHg to 5.99mmHg to a post-calibration range of 0.14mmHg to 0.48mmHg. A significant improvement in measurement accuracy was noted. Using this calibrated simulator, the "Accurate 24" cuffless continuous BP monitor was adjusted.

Conclusion: This research primarily revolves around the accurate measurement of pulse waves, which is of paramount importance to the precision of cuffless BP monitors. By leveraging hemodynamic characteristics, we designed a pulse wave simulator and identified a standard performance validation method. Initially, the cuffless BP monitor under study is modeled using MLR with the pulse wave simulator, following which its characteristics and discrepancies are measured against the simulator. The simulator designed under this research principle is suitable for mass production and can quickly validate the features of cuffless BP monitors. It holds potential as an essential component of product inspection in the industrial production phase, especially as cuffless BP monitors become ubiquitous in the future.