摘要
具診斷等級之行動健康照護平台，旨在為現代人建立一個便利的遠距照護環境。雖然行動醫療已不是一個創新的研究課題，但是新興的穿戴感測技術和時下普及的行動設備卻可以使這樣的系統變得比以前更加完善且強大。此外，醫生和工程師之間，跨領域的合作更是另一項使這些研究在過去幾年變得更加完善的推動力。 在我們的平台上，我們利用自行開發的心電圖記錄原型模組來記錄用戶的單導極（Ⅱ導極）心電圖信號。原型的整體重量約35克（包含一顆850毫安時的充電鋰電池），且大小只有一般信用卡的一半。我們所開發的智慧手機應用程式擁有良好且人性化的使用者介面，它可以讓我們利用現今廣泛流行的行動裝置做為整體平台的中繼站，協調可穿戴式量測器和雲端伺服器之間的溝通與資料傳遞。由於當今手機的計算能力以及可程式化特性，我們所提供的手機應用程式不僅僅可以將所記錄的數據轉送到醫療雲，也可以為用戶們提供一些具有價值的服務和功能。這正是我們平台的主要值之一。 雖然我們的原型模組和應用程式可以提供許多功能，但它仍然擁有令人稱羨的低功耗特性。對於無線量測的原型來說，一顆850毫安時的電池可以讓它連續地記錄並傳送數據至少24小時（整個傳感器的總功耗約為145毫瓦，而其中約110毫瓦的功耗則是來自於所使用的藍芽2.1模組。當我們更換為4.0版本的藍芽低功耗（BLE）模組後，整體系統將會擁有更低的功耗）。對於手機應用軟體來說，我們對程式中所有的服務線程都進行了最佳化的工作，以為手機帶來最長的電池壽命。根據我們的實驗，一支帶有1800毫安時電池的四核心Android手機可以在執行我們的程式時擁有至少6.5小時的電池壽命。除此之外，高端的硬體規格在我們所設計的軟體上並非必要。一支具有800 MHz單核心處理器的Android手機就足以非常流暢的運行我們所設計的應用程式。 雲端醫療是我們平台的另一主要價值。這樣的雲端服務不僅可以提供準確的檢測方法，而且還提供先進生物醫學信號處理的分析技術。在我們的平台中，我們設計了一個精確的演算法來檢測所記錄到的單導極心電圖信號。經過MIT-BIH心律不整數據庫的驗證，我們的演算法具有著強大的效能。靈敏度（Se%）和預測能力（+P％）分別可以達到98.80％和99.19％。如果我們忽略MIT-BIH數據庫中極端的幾種情況：108，203，222，檢測結果將會高達99.45％和99.42％。除此之外，許多其他心電圖的分析方法也在我們的平台上被執行。而所有這些方法都已在不同的心臟分析研究中被證明是具有意義的。 我們同時也利用所實現的分析方法和時頻分析技術進行了另一個創新研究。我們希望能找出非酒精性脂肪肝（NAFLD）患者和正常人之間，心率變異性（HRV）指標的顯著差異。這可以幫助我們確定人體的肝臟和心臟之間是否有一些相互影響。果然，此一研究結果表明，非酒精性脂肪肝與ln sdNN的減少以及0V百分比的增加有關。因此我們認為，透過HRV參數對自主神經功能紊亂做進一步的風險分層對於NAFLD患者是必要的。
並列摘要
In this thesis, a diagnostic grade mobile healthcare platform is designed to establish a mobile telecare environment for modem people. Although mHealth is not an innovative research topic, the improvements of wearable sensing technologies and today’s mobile devices can make such systems become more consummate and powerful than before. Additionally, the cooperation between physicians and engineers is another main driving force for these researches to become even more consummate in the past few years. In our platform, a wearable ECG recording prototype is developed to record single-lead (lead II) ECG signals of users. The whole prototype weight about 35 g (including an 850 mAh battery), and is only half of a credit card in size. With our well-developed and user-friendly smartphone application program, we can use a widespread mobile device as a hub to coordinate connections between wearable sensors and cloud techniques. Due to the computing power and the programability of today’s mobile phone, our hub can not only transmit recorded data onto medical cloud, but also provide several useful services and functions for users. This also is one of the main values of our platform. Although our prototype and application can provide such many functions, it is still amazingly energy efficient. For the sensor node prototype, an 850 mAh battery can make such device continuously recording and transmitting data for at least 24 hours (the total power consumption of the whole sensor is about 145 mW, while about 110 mW for the wireless bluetooth 2.1 module only). This is expected to be much longer after replacing the bluetooth module with the 4.0 Bluetooth low energy (BLE) version. For the mobile application, all service threads of the program are optimized to bring longest battery life to mobile devices. However, the actuarial battery life of the mobile hub depends on the specification of different devices. Generally, a quad-core Android phone with a 1800 mAh battery can have at least 6.5 hours battery life according to our experiments. Besides of this, high-end hardware is not required at our mobile hub. A general Android phone with 800 MHz single core CPU can run our program smoothly as well. Besides of the mobile hub, the patient-centric medical cloud is another priceless part of our mobile healthcare platform. Such cloud service can provide not only accurate detecting algorithms, but also advanced analyzing methods for biomedical signal processing. We designed an accurate algorithm to detect R-R intervals of recorded single-lead ECG signals using different operation method (DOM). Also, this algorithm is verified with MIT-BIH arrhythmia database, and the results show our algorithm has great performance. The sensitivity (Se%) and positive predictivity (+P%) of our DOM algorithm can achieve 98.80% and 99.19%, respectively. And if we ignore 3 worst cases of MIT-BIH database: 108, 203, 222, as many other researches do, we can even higher the value to 99.45% and 99.42%. In addition to this, many other ECG analyzing methods are also implemented in our platform. All of these methods were proved to be useful in different cardiac analysis. We also use the implemented analyzing methods and some time-frequency analysis methods in another innovative study. In this study, we intend to find the significant difference between heart rate variability (HRV) indices analyzed from non-alcoholic fatty liver decease (NAFLD) patients and normal people. This can help us define the influence of liver on heart in human body. Expectedly, results of this study shows that, NAFLD is associated with decreased ln sdNN and increased 0V percentage. Therefore, we declare that, further risk stratification of autonomic dysfunction with falls or cardiovascular diseases by these HRV parameters is required in patients with NAFLD.
參考文獻
H.-P. Ma, Y.-L. Ho, and C.-H. Chen, “Influence of non-alcoholic fatty liver disease
June 2012.
[2] J. E. Hall, Guyton and Hall textbook of medical physiology, 12nd ed. Philadelphia,
[6] R. M. Rangayyan, Biomedical Signal Analysis: A Case-Study Approach, 1st ed. Piscataway,
NJ, USA: Wiley-IEEE Press, December 2001.