現代人因為工作的關係常常過著不規律的生活作息、不均衡的飲食習慣，累積血管中的脂肪和膽固醇造成動脈粥狀硬化，隨著年齡老化加劇動脈硬化的狀況，進而增加中風、心臟病、高血壓等疾病的風險，這些疾病都有突然發生的風險。目前能夠檢測動脈硬化的儀器通常都需要專業人員協助進行操作與量測，量測的場地空間也容易受限制，若能有容易穿戴且能在居家環境中隨時量測的儀器協助檢測，配合機器學習將數據做進一步分析應用，便能有效且及早預防上述疾病的發生。
本研究使用本實驗室自行開發的穿戴式脈波量測系統進行動脈硬化數據的深入探討分析，量測生理訊號包括心電訊號描述放大器(Electrocardiography, ECG)、動脈血壓波形感測器(Blood Pressure Wave, BPW)及血管光容積感測器(Photoplethysmography, PPG)，並與雙和醫院合作，在土城社區場域收案，為了符合社區收案情況因此對穿戴式裝置進行改良，增加一般民眾在穿戴上的親和力，且整理出社區民眾針對整體系統裝置的問題回饋，最後進一步將整體收案系統體積縮小，並予以實現。
在社區收案完成後，將社區民眾的生理數據進行頻域分析計算出生理參數，配合收案團隊的動脈硬化儀與心音儀進行數據比較，觀察出動脈血管特性與生理參數的關聯性，結合機器學習的八種演算法，建立最適合的機器學習模型，將其應用在分類血管老化的數據上。
實驗結果表示量測到的BPW數據在頻域上正常受試者與動脈硬化受試者有明顯差異，並觀察出在不同年齡層的數據會因為年齡老化的狀況不同而有影響，因此透過年齡的區分統計觀察出數據差異，並建立機器學習模型，找出具有最好分類能力的演算法，最後再進行每筆數據實際測試，其中多層感知器(Multilayer perceptron, MLP)演算法與隨機森林(Random forest, RF)演算法都具有很好的分類能力，代表此穿戴式系統結合機器學習模型能很好應用在社區量測數據上，而心音儀數據也在觀察完統計結果後，利用數據差異最為明顯的心音指標進行機器學習訓練，線性區別分析(Linear Discriminant Analysis, LDA)演算法也具有一定程度的分類能力。
本實驗室開發之裝置經過改良與結合機器學習進一步應用能更好的在社區居家環境中使用，且具有一定程度的解析能力，在臨床應用上能夠隨時監控使用者的生理數據，協助醫師在社區居家環境達到早期偵測與早期預防的目標。

Modern people often lead irregular lives and unbalanced eating habits due to work. The accumulation of fat and cholesterol in blood vessels causes atherosclerosis. As aging ages, the condition of arteriosclerosis increases, which also increases the risk of strokes, heart diseases, Hypertension and other diseases, these diseases all have the risk of sudden occurrence. The current instruments that can detect arteriosclerosis usually require professional assistance in operation and measurement, and the space of the measurement site is also easily limited. If there are instruments that are easy to wear and can be measured at any time in the community environment to assist in the detection, cooperate with the machine Learning to further analyze and apply the data can effectively to early prevent the occurrence of the above-mentioned diseases.
In this study, a wearable pulse wave measurement system developed by our laboratory was used to analysis of arteriosclerosis data. The measured physiological signals include electrocardiography (Electrocardiography, ECG) , arterial blood pressure waveform sensor (Blood Pressure). Wave, BPW) and Photoplethysmography (PPG). And cooperated with Shuanghe Hospital to accept cases in the Tucheng community. In order to meet the acceptance situation in the community, the wearable device was improved to increase the general public Affinity in wearing. Then sorted out the community people’s feedback on the problems of the overall system device. Finally further reduced the size of the overall collection system and realized it.
After the completion of the community acceptance, the physiological data of the community people are analyzed in the frequency domain to calculate the physiological parameters, the arteriosclerosis instrument and the heart phonograph of the acceptance team are used to compare the data. The correlation between the arterial vascular characteristics and the physiological parameters is observed. The eight algorithms of machine learning establish the most suitable machine learning model and apply it to the classification of vascular aging data
The experimental results show that the measured BPW data is significantly different in the frequency domain between normal subjects and arteriosclerosis subjects, and it has been observed that the data at different age groups will be affected by different conditions of aging. Through the age we can observe the data difference by distinguishing statistics, and establish a machine learning model to find out the algorithm with the best classification ability, then perform the actual test on each data. Among them, the MLP algorithm and the RF algorithm have good classification capabilities. This means that this wearable system combined with machine learning models can be well applied to community measurement data. After observing the statistical results, the heart sound indicators with the most obvious data differences are used for machine learning training. The LDA algorithm also has a certain degree of classification ability.
The device developed in this laboratory can be better used in the community home environment after improvement and combined with machine learning having a certain degree of analytical ability. It can monitor the physiological data of the user at any time in clinical application and assist physicians in the community environment. And achieves the goal of early detection and early prevention.

[1] 彰化基督教醫院血管醫學防治中心與血管暨基因體研究中心.(2010). 血管醫學臨床與研究電子報第19期. Retrieved from https://www.cch.org.tw/vmpc/news/news_detail.aspx?oid=141&no=7
[2] 康健雜誌 特刊第26期.(2012). Retrieved from https://www.commonhealth.com.tw/article/article.action?nid=72096&from=search
[3] Lusis A. J. (2000). Atherosclerosis. Nature, 407(6801), 233–241. Retrieved from https://doi.org/10.1038/35025203
[4] Alexander, R. W. (1995). Hypertension and the pathogenesis of atherosclerosis: oxidative stress and the mediation of arterial inflammatory response: a new perspective. Hypertension, 25(2), 155-161.
[5] 衛生福利部統計處.(2019). 108年國人死因統計結果. Retrieved from https://dep.mohw.gov.tw/DOS/cp-4927-54466-113.html
[6] Kramer, C. M., & Anderson, J. D. (2007). MRI of atherosclerosis: diagnosis and monitoring therapy. Expert review of cardiovascular therapy, 5(1), 69-80.
[7] Soeki, T., & Sata, M. (2016). Inflammatory biomarkers and atherosclerosis. International heart journal, 15-346.
[8] Yambe, T., Yoshizawa, M., Saijo, Y., Yamaguchi, T., Shibata, M., Konno, S., ... & Kuwayama, T. (2004). Brachio-ankle pulse wave velocity and cardio-ankle vascular index(CAVI). Biomedicine & pharmacotherapy, 58, S95-S98.
[9] Shirai, K., Utino, J., Otsuka, K., & Takata, M. (2006). A novel blood pressure-independent arterial wall stiffness parameter; cardio-ankle vascular index (CAVI). Journal of atherosclerosis and thrombosis, 13(2), 101-107.
[10] Namekata, Tsukasa & Suzuki, Kenji & Ishizuka, Norio & Shirai, Kohji. (2011). Establishing baseline criteria of cardio-ankle vascular index as a new indicator of arteriosclerosis: A cross-sectional study. BMC cardiovascular disorders. 11. 51. 10.1186/1471-2261-11-51.
[11] Dillier, R., Kobza, R., Erne, S., Zuber, M., Arand, P., & Erne, P. (2011). Noninvasive detection of left-ventricular systolic dysfunction by acoustic cardiography in atrial fibrillation. Cardiology research and practice, 2011.
[12] 麗臺科技官方網站 Retrieved from https://www.leadtek.com/cht/products/telecare(23)/%E3%80%8C%E5%81%B5%E8%B2%BC%E5%BF%83%E3%80%8D%E5%BF%83%E9%9B%BB%E5%BF%83%E9%9F%B3%E7%B4%80%E9%8C%84%E5%99%A8(732)/detail
[13] 宋巧娟（2016）。循環量測指標應用於冠狀動脈狹窄疾病之研究。國立臺灣科技大學醫學工程研究所碩士論文，台北市。
[14] 許家良（2014）。血壓波形的諧波分析應用於多囊性卵巢及中風之偵測及療效評估研究。國立臺灣科技大學應用科技研究所博士論文，台北市。
[15] 劉人銘（2018）。循環量測應用於服用高血壓藥物患者之研究。國立臺灣科技大學醫學工程研究所碩士論文，台北市。
[16] 陳均曄（2019）。非侵入式循環訊號量測系統應用於不同癌症之偵測與評估。國立臺灣科技大學醫學工程研究所碩士論文，台北市。
[17] 楊舒涵（2015）。都卜勒微循環血流與動脈血壓波形指標 應用於代謝症候群之研究。國立臺灣科技大學醫學工程研究所碩士論文，台北市。
[18] 陳冠彰（2016）。循環訊號量測應用於乳癌患者病程監控之研究。
[19] Li, B., Gao, H., Li, X., Liu, Y., & Wang, M. (2006). Correlation between brachial-ankle pulse wave velocity and arterial compliance and cardiovascular risk factors in elderly patients with arteriosclerosis. Hypertension research : official journal of the Japanese Society of Hypertension, 29(5), 309–314. https://doi.org/10.1291/hypres.29.309
[20] Ungvari, Z., Tarantini, S., Donato, A. J., Galvan, V., & Csiszar, A. (2018). Mechanisms of vascular aging. Circulation research, 123(7), 849-867.
[21] CHUDUC, Hoang; NGUYENPHAN, Kien; NGUYENVIET, Dung. A review of heart rate variability and its applications. APCBEE procedia, 2013, 7: 80-85.
[22] LEE, Hae-Young; OH, Byung-Hee. Aging and arterial stiffness. Circulation Journal, 2010, 1010120923-1010120923.
[23] SHIRWANY, Najeeb A.; ZOU, Ming-hui. Arterial stiffness: a brief review. Acta Pharmacologica Sinica, 2010, 31.10: 1267-1276.
[24] 林育楷.(2019). 中國醫訊第187期-動脈硬了 問題來了 血管裡的風暴.Retrieved from https://www.cmuh.cmu.edu.tw/NewsInfo/NewsArticle?no=4672