周邊動脈疾病（PAD）是一種由動脈粥狀硬化引起的血管疾病，常合併其他心血管問題。然而， PAD 初期多無明顯症狀，臨床診斷常遭延誤。目前標準檢查為踝臂血壓比值（ABI），但在糖尿病、慢性腎病族群中準確度較低，且應用場域受限。本研究建置一套以四肢 PPG 與 ECG 訊號為基礎的 PAD 評估系統，整合自製多通道感測模組與訊號擷取裝置，能同步量測四肢生理訊號。考量脈波傳導時間（PTT）需依賴穩定訊號方可準確計算，進一步開發結合短時傅立葉轉換（STFT）與卷積神經網路（VGG-16）架構的同步訊號篩選輕量模型（SSS-Model），以自動辨識穩定訊號片段。結果顯示， SSS-Model 處理兩分鐘多通道訊號僅需 0.067 秒，分類準確度達 86%，F1-score 為 0.81，展現即時應用潛力。統計分析發現腳趾 PTT 與 ABI 呈顯著負相關（r = –0.88），健康與 PAD 族群間之 PTT 差異具統計顯著性（p < 0.005）。此外，本系統有效辨識 ABI 正常但有 PAD 臨床症狀的患者與健康族群（p < 0.05），突破 ABI 在評估特定族群患有 PAD 的限制。此系統具準確性高、操作簡便與環境適應性佳等優勢，未來可應用於居家、初級醫療與長照機構等場域，作為 PAD 初步篩檢工具。

Peripheral artery disease (PAD) is a vascular condition caused by atherosclerosis and is often associated with other cardiovascular problems. However, PAD is typically asymptomatic in its early stages, leading to delayed clinical diagnosis. The current standard diagnostic method is the ankle-brachial index (ABI), but its accuracy is reduced in populations such as those with diabetes or chronic kidney disease, and its application is limited to specific clinical settings. In this study, we developed a PAD assessment system based on four-limb PPG and ECG signals. The system integrates a custom-designed multi-channel sensing module and signal acquisition device capable of synchronously recording physiological signals from all four limbs. Since pulse transit time (PTT) requires stable signals for accurate calculation, we further developed a lightweight simultaneous signal selection model (SSS-Model), combining Short-Time Fourier Transform (STFT) and a convolutional neural network based on the VGG-16 architecture to automatically identify stable signal segments. Results showed that the SSS-Model processed two-minute multi-channel recordings in just 0.067 seconds, achieving a classification accuracy of 86% and an F1-score of 0.81, demonstrating strong potential for real-time applications. Statistical analysis revealed a significant negative correlation between toe PTT and ABI (r = –0.88), and significant differences in PTT between the healthy and PAD groups (p < 0.005). Moreover, the system successfully distinguished patients with normal ABI but clinical symptoms of PAD from healthy individuals (p < 0.05), overcoming the limitations of ABI in detecting PAD in specific populations. With advantages such as high accuracy, user-friendly operation, and environmental adaptability, this system holds promise as a preliminary PAD screening tool for home use, primary care, and long-term care settings.