本研究啟發於瘀傷年齡判定之臨床需求。由於瘀傷內部的物質如血紅蛋白會隨時間推移產生分解與濃度變化，因此若能精確定量其濃度，將有助於瘀傷年齡判定。本論文旨在開發並優化一套光聲二維掃描影像系統(Photoacoustic Imaging System)，以實現皮下血紅蛋白濃度的定量。 為克服傳統穿透式光聲系統需將樣本浸於水中、適用厚度受限且對位穩定度不足等缺點，本系統導入二維掃描振鏡與光聲結合器(Opto-ultrasonic combiner)，建構出一套同軸光聲掃描影像系統。硬體整合LabVIEW平台進行訊號擷取與同步控制，完成10，000像素之掃描僅需40秒，並透過實驗驗證了系統的空間解析度與基礎光聲物理特性。 在定量分析方面，本研究利用MATLAB建立高斯過程迴歸(Gaussian process regression)之機器學習模型。實驗中利用532 nm脈衝雷射作為激發光源，樣本則採用Ponceau 4R色素與皮膚仿體模擬皮下血管，將影像強度、樣本深度、雷射功率及功率-強度擬合斜率等多維參數導入模型訓練。隨後利用已知濃度之樣本進行測試，驗證預測準確性並剖析誤差來源。此外，為進一步驗證臨床應用潛力，本研究亦對新鮮雞心進行活體外掃描，並將血管影像特徵匯入模型獲得濃度預測分布圖。最後針對研究過程中遇到的問題制定系統未來開發上的目標。;Inspired by the clinical demand for bruise age determination， this study focuses on the precise quantification of subcutaneous hemoglobin. Since substances within bruises， such as hemoglobin， degrade and change in concentration over time， accurately quantifying these variations can facilitate the determination of a bruise′s age. This thesis aims to develop and optimize a 2D photoacoustic (PA) scanning imaging system to achieve subcutaneous hemoglobin quantification. To overcome the limitations of traditional transmission-mode PA systems—such as sample immersion requirements， restricted applicable thickness， and poor alignment stability—a 2D scanning galvanometer and an opto-ultrasonic combiner were integrated to construct a coaxial PA scanning imaging system. Controlled via a LabVIEW platform for synchronized signal acquisition， the system achieves a 10，000-pixel high-resolution scan in merely 40 seconds. Its spatial resolution and fundamental physical characteristics were experimentally verified. For quantitative analysis， a Gaussian Process Regression (GPR) machine learning model was developed in MATLAB. Utilizing a 532-nm pulsed laser as the excitation source， tissue-mimicking skin phantoms and Ponceau 4R dye were used to simulate subcutaneous blood vessels. Multidimensional parameters—including image intensity， sample depth， laser power， and power-intensity fitting slopes—were fed into the model for training. The prediction accuracy was subsequently validated using samples of known concentrations， followed by an analysis of potential error sources. To further demonstrate clinical potential， ex vivo scans of fresh chicken hearts were performed， successfully predicting vascular concentrations via the GPR model. Finally， based on the challenges encountered during the research， future development goals for the system are outlined.
