生醫阻抗量測系統之設計與開發

Abstract

生醫阻抗分析（bioelectrical impedance analysis, BIA）是一種用來判斷生物材料（biological materials）組成結構的非侵入性方法，雖然已有多組研究團隊利用生醫阻抗分析技術，但目前研究多半依賴成本甚高的商業儀器（commercial instrumentation）來完成。為了達到降低檢驗成本的目標，本研究以德州儀器（Texas Instruments, TI）專用於生醫阻抗量測的積體電路解決方案AFE4300為核心，提出全套定制硬體系統的設計與開發，方便以定點照護檢驗（Point-of-Care Testing, PoCT）的方式進行生醫阻抗的量測。本研究開發的系統提供相當於商業儀器的量測功能，但硬體總成本僅為八十美元（約新臺幣兩千三百六十元）。再來，本研究利用此系統從受試者的肱二頭肌（biceps brachii muscle）收集不同肌肉狀態的數據。此外，本研究也包括機器學習（Machine Learning, ML）應用程式的開發，以接近百分之八十的準確率辨別四種不同的肌肉狀態，若僅對於兩種肌肉狀態加以辨別，該應用程式的準確率甚至高達百分之九十以上。

Bioelectrical Impedance Analysis (BIA) is a non-invasive method for estimating the composition of biological materials. Several research groups have come to use bioimpedance measurements in their studies, mostly relying on commercial – and therefore high-cost – instrumentation. In this research, a fully custom-built low-cost device capable of performing bioimpedance measurements in a Point-of-Care Testing (PoCT) environment was designed and implemented using the AFE4300 from Texas Instruments as a specialized integrated circuit (IC) solution. The system offers comparable measurement features to commercial instrumentation, but at a hardware cost of only USD 80 (approximately NTD 2,360) in total. The system was then used to collect muscle state data from the biceps brachii muscle of subjects for different muscle states. This work also includes the development of a machine learning application that succeeded in distinguishing between four different muscle states with an accuracy of almost 80%. When only distinguishing between two different muscle states, the accuracy even raised to more than 90%.