| 摘要: | 過去幾年全球在新冠疫情的席捲下，遠距醫療(Telemedicine)的發展逐漸蓬勃，透過居家設備獲取人體生理訊號並分析成為發展重點，現今市面上的醫療檢測儀器有物體體積較大以及價格昂貴的問題，透過影像提取是最為便利以及低成本的方式，非接觸式心衝擊波(Ballistocardiogram)為其中代表技術，形成原因為心臟每博跳動輸出血液，衝擊血管壁的反作用力造成人體質心位移，這些週期性微小的位移量可以透過影像技術提取，隱含心臟收縮時的心血管功能資訊，具有遠端診斷心臟功能正常與否的潛力，本研究專注在心臟衰竭病人(Heart failure, HF)上，其心臟泵血功能降低導致心輸出量(Cardiac output, CO)減少，造成波形產生變化，心衝擊波波形上的特徵點與正常狀態下出現差異，量化差異可以協助即時判斷心臟衰竭程度，本研究先透過心衝擊波閉迴路數學模型模擬心臟收縮力下降時，心衝擊波的波形變化，再進行實體模擬實驗，利用影像方式收集10位正常人40s正常呼吸與30s伐式呼吸(Valsalva maneuver)之資料，伐式呼吸造成靜脈回流量減少用以模擬心輸出量降低之情形，最後收集臨床心臟衰竭患者的資料作驗證;本研究方法透過影像利用Kanade-Lucas-Tomasi(KLT)光流法追蹤感興趣區域(ROI)的特徵點，根據每個影像禎中的運動軌跡提取出心衝擊波，並利用波形函數(Wave-shape function)對生理訊號進行分析，波形函數分解是一種具抗噪、分割週期優勢的訊號處理方法，尤其在非接觸式心衝擊波以及易受運動干擾、噪音影響之下，波形函數是很好的抗噪重構方法，並且使重構的單一週期心衝擊波不受心率影響，使計算的特徵間期更具穩健性，並將特徵間期與超音波射血率作統計分析，以及與射血前期(Pre-ejection period, PEP)、等容收縮期(Isovolumetric Contraction Time, IVCT)交互驗證，證明使用心衝擊波分析能判斷心臟功能是否異常以實現居家心衰檢測及時系統。;In recent years, under the sweeping impact of the COVID-19 pandemic, the development of telemedicine has grown rapidly. Acquiring and analyzing physiological signals through home-based devices has become a major focus. However, existing medical diagnostic instruments are often bulky and expensive. Image-based signal extraction offers a convenient and low-cost alternative. Among the representative non-contact technologies is Ballistocardiography (BCG), which originates from the recoil movement of the human body caused by blood being forcefully ejected from the heart and impacting the vascular walls during each cardiac cycle. These periodic micro-movements can be extracted via imaging techniques and contain valuable cardiovascular information related to cardiac contraction, showing great potential for remote diagnosis of cardiac function. This study focuses on patients with heart failure (HF), a condition characterized by decreased cardiac pumping efficiency and reduced cardiac output (CO), which leads to alterations in the BCG waveform. The feature points on the BCG waveform differ from those in normal conditions, and quantifying these differences may aid in real-time assessment of heart failure severity. In this study, we first use a closed-loop mathematical model of BCG to simulate waveform changes under reduced cardiac contractility. We then conduct physical simulation experiments by collecting video data from 10 healthy subjects under two conditions: 40 seconds of normal breathing and 30 seconds of Valsalva maneuver. The Valsalva maneuver reduces venous return and is used to mimic reduced cardiac output. Finally, we validate the method with clinical data from heart failure patients. Our proposed method uses the Kanade-Lucas-Tomasi (KLT) optical flow algorithm to track feature points in regions of interest (ROI) within video frames. The motion trajectories in each frame are used to extract BCG signals. Wave-shape function analysis is then applied to these physiological signals. Wave-shape function decomposition is a noise-robust method with strong performance in cycle segmentation, especially under the non-contact BCG setting where signals are prone to motion artifacts and noise. This method reconstructs single-cycle BCG waveforms independently of heart rate, enhancing the robustness of the computed interval features. Finally, we perform statistical analysis comparing the extracted features with echocardiographic ejection fraction (EF) values, and cross-validate with pre-ejection period (PEP) and isovolumetric contraction time (IVCT). The results confirm that BCG-based waveform analysis is effective in identifying abnormal cardiac function, enabling the realization of a real-time, home-based heart failure monitoring system. |
