| 摘要: | 近年來健康問題受到世界許多研究人員的重視，非侵入式的測量方法也開
始被應用於健康偵測上。雷射都普勒血流儀 (Laser Doppler flowmetry, LDF)
其原理是利用都普勒效應 (Doppler effect)，被運用於測量組織微循環血流
流速的技術。 LDF 是一種非侵入式的測量裝置，許多研究人員利用 LDF 做
醫學上的研究，發現微循環與身體狀況息息相關，所以 LDF 適合應用於健
康偵測上的領域，但 LDF 機器成本昂貴，且取得不易，導致相關應用並不
普及，本論文延續前人的研究成果，以市面上可購買的電子材料建構一台
實際可用的 LDF 原型機，並且更進一步，設計一套精準的數學迴歸分析模
型，用來校正原型機所測量到的信號特徵，希冀其能夠與商用 LDF 機器所
量到的特徵完全一致。
關於信號特徵之驗證與校正，在本論文中，我們以食指中心作為信號
量測點，將商用機與原型機之血流訊號分別在信號層面與特徵層面做詳細
的比對與分析，以期建立模型並修正原型機的誤差。實驗分析結果：在原
始訊號上商用機與原型機波形沒有關聯性，藉由 ECG(心電圖) 切割後，可
以發現波形的相關性，其原因可能因為不是同時測量，微循環訊號可能有
差距，在平均整體結果後，可以看出手指微循環的整體趨勢。而在生理特
徵上，在未校正前相關性約在 0.7 上下，可以解釋兩者存在相關性，但有
訊號上的誤差，在由校正演算法一校正後，結果發現兩者特徵的相關性開
始往正相關 1.0 逼近，進一步使用演算法二校正後，結果顯示，最好的結
果已經可以逼近於 0.9999，已經趨近於完全正相關。最後整體評估結果顯
示，本論文所提出的方法，可以驗證 LDF 原型機的正確性，並可以在校正
後得到與商用機極高的相關性。
在本論文的研究成果可以幫助 LDF 原型機進入商用化的階段。由於
LDF 原型機的成本價格只有五萬元左右，在大規模生產後預估成本可以進
一步下降。此價格與目前市面上 LDF 的商用機相比具有巨大的競爭優勢，
經由本論文的迴歸模型校正之後，所有基於 LDF 商用機所設計出來的應用
場合，都可以直接套用在本論文所提出的 LDF 原型機。故此，本論文的研
究成果，可以大幅降低應用 LDF 的成本，加速其大規模的普及性應用於各
大醫療院所、小診所、健身中心等等。另外一個具有未來性的應用為穿戴
式裝置。本論文所設計的 LDF 原型機，可以用積體電路的方式，縮小其尺
寸成為單一 IC。若能順利實作成功，將來可以將此 LDF IC 內嵌到穿戴式
裝置上，例如：智慧手錶、智慧眼鏡、智慧衣、智慧褲等等。如此將能使
得 LDF 的應用更加普及，並且對行動照護、遠端醫療等高齡化社會所關注
的主題產生正面的貢獻。;In recent years, many researchers have paid attention to health problem in the
world. Non-invasive measurement methods were applied to detect health prob
lems. Laser Doppler flowmetry (LDF), a non-invasive method, is used to measure
red blood cells in a tissue. Researchers used LDF to do research in medicine,
and they found that the microcirculation and health condition are closely related.
Therefore, LDF is very applicable to healthcare. However, the commercial LDF
machines are very expensive and hard to obtain, which lead to very few relevant
application and popularization. In this thesis, continuing previous researches, a
LDF Prototype was built by electronic materials which can be purchased com
mercially; Further, we design an accurate mathematical regression analysis model
to calibrate feature of LDF prototype, and we hope to match with business LDF
feature.
On verification and calibration of the LDF signal feature, we compared the
data obtained by business LDF model and LDF prototype to establish a model
for calibrating error. We took the center of forefinger as the measure point and
measured the signal layer and feature layer. The result of the experiment showed
that before calibration, the correlation coefficient of physiological feature is about
0.7, and after we used our Regression Model to calibrate physiological feature, the
correlation coefficient reached nearly 0.9999, which was close to a perfect positive
correlation. The overall evaluation results showed that the proposed method can
verify and insure the correctness of the LDF prototype. Also, LDF prototype can
obtain high correlation with business LDF after calibration.
The results of this thesis that can help LDF prototype into the commercial
ization stage. Since the LDF prototype only cost about fifty thousand NTD, and
the estimated cost can be further reduced after the mass production. This price is
highly competitive comparing to business LDF. Through regression model to cal
ibrate it in this thesis, the applications of all business LDF can be directly applied
in the LDF prototype. Therefore, the result of this thesis, the LDF prototype can
significantly reduce the cost, and promote popularization to use in medical insti
tutions, fitness centers, etc. In addition, the application on the wearable device is
also promising. The LDF prototype can be built into integrated circuits that can
reduce the size to a single IC. If it is successfully implemented, the LDF prototype
IC can be plugged in wearable device. For instance, smart glasses, smart watch,
smart clothing and smart pants, etc. The LDF application can be more popular
ized, and it has a positive contribution to the action care, remote medical, aging
society. |
