| 摘要: | 中文摘要
背景：周邊動脈阻塞PAOD）是一個主要的公共衛生負擔，需要更深入的人群篩查。它是低收入國家最被低估的疾病之一，可能導致嚴重的心血管事件。因此，PAOD 的早期檢測對於臨床實踐至關重要。 PAOD 傳統上通過都普勒超音波和踝臂指數 (ABI) 來識別，踝臂指數 (ABI) 源自手臂和踝袖套記錄的最大壓力波形的比率，可以是一種方便的替代方法。然而，它需要訓練有素的操作員採用嚴格的方法，而且也很耗時。而使用傳統的ABI值，由於ABI值落在正常範圍（0.9至1.4），醫生不容易識別小腿動脈的閉塞情況。
目的：光體積變化 (PPG) 與壓力波形類似，但價格較低，可以從手指或腳趾記錄，並測量動脈內脈壓傳播引起的血液體積變化。本研究的目的是驗證與傳統 ABI 相比，使用源自 PPG 的特徵可以幫助我們更好地識別 PAOD，並且可以用作 PAOD 檢測的替代工具。而且，獲得 PPG 特徵，與傳統 ABI 在正常範圍內的其他位置相比，可以更好地區分腿部阻塞的情形。
材料與方法：共招募130例疑似PAOD患者，採用多普勒超聲方法診斷PAOD，並使用壓力袖帶計算傳統ABI。然後從4個位置（左右食指和腳趾）同時記錄PPG信號10秒，然後使用單個探針從相同位置記錄四個10秒連續的PPG信號。 PPG 信號的噪聲通過帶通濾波器 (1-10 Hz) 去除，並且在峰值檢測後，脈動波形的峰值被對齊並平均。平均波形用於從信號的二階導數中提取波形起點、峰值、拐點和終點。從 PPG 中提取多個特徵並將其用於分類。發現 PPG 特徵的 AUC 為 0.8481，敏感性為 74%，特異性為 84%，而傳統 ABI 的 AUC 為 0.7914，敏感性為 52%，特異性為 98%。發現與傳統 ABI 相比，PPG 特徵具有更好的靈敏度和 AUC。因此，為了改善結果，從信號中提取更多特徵並進行分類。發現在最佳閾值下，PPG 特徵的 AUC 為 0.8448，敏感性為 73%，特異性為 84%，而傳統 ABI 為 0.7692 AUC，敏感性為 48%，特異性為 98%。與傳統的 ABI 相比，使用 PPG 特徵對動脈閉塞位置進行分類給出了更好的結果。 PPG 特徵可以識別 91.6% 的膝蓋以上患者，65% 的膝蓋以下患者，而在最佳閾值 0.9 下，ABI 可以識別 75% 的膝蓋以上患者，36.7% 的膝蓋以下患者。然後脈搏高度 90% 處的峰度和舒張壓可以更好地區分膝蓋上方和下方區域的動脈。
討論：與傳統的 ABI 相比，PPG 特徵表現出更好的靈敏度，AUC 更大。但與傳統 ABI 相比，PPG 特徵的特異性較低。該系統的一個優點是，與之前的研究相比，我們使用了來自腳趾的 PPG 信號，並使用臨床標準超音波確診的結果進行分類，這比與 ABI 值範圍進行比較更可靠。在上述區域的數據量最少的情況下，峰度在閾值 1.7208 處提供了更好的區分，而舒張寬度為膝蓋以下的動脈提供了更好的結果。因為從 PAOD 患者獲得的近 30% 的 PPG 信號由於閉塞的嚴重程度而存在雜訊，並且無法從這些值中獲得任何有用的信息。由於阻塞較嚴重時無法獲取PPG信號。因此，它可以用作 PAOD 檢測初步研究的工具，無需複雜的測量技術。當我們根據阻塞位置（例如膝蓋上方和下方的動脈）對患者進行分類時，發現 PPG 特徵的最佳閾值比傳統 ABI 提供了更好的識別結果。因此，與傳統的 ABI 相比，PPG 可用作識別所有動脈位置阻塞的初步調查工具，給出了有希望的結果。
結合深度學習技術可以提供潛在的改進，為更準確、更高效的周邊動脈阻塞診斷和治療計劃鋪平道路，最終有利於患者的治療結果和醫療保健實踐。
;English Abstract
Background: Peripheral Arterial Occlusive disease (PAOD) is a major public health burden requiring more intensive population screening. It is one of the most underrated disease in the low-income countries and it could lead to severe cardiovascular events. Early detection of PAOD is therefore crucial for clinical practice. The PAOD is identified traditionally through the Doppler ultrasound and Ankle brachial index (ABI) which derived from the ratio of maximal pressure waveforms record from arm and ankle cuffs can be a convenient alternative method. However, it requires a rigorous methodology by trained operators and also time-consuming. And by using the traditional ABI value it is not easy for the doctors to identify the occlusion on the lower leg arteries due to the ABI values falling in the normal range of (0.9 to 1.4).
Objective: Photoplethysmography (PPG), which is similar to the pressure waveform but less expensive, can be recorded from the fingers or toes and measures volumetric blood changes brought on by pulse pressure propagation within arteries. The purpose of this study is to validate the ability of using the features derived from the PPG can helps us to identify the PAOD better when compared to the traditional ABI and can be used as alternative tool for PAOD detection.
And also, to get the PPG feature which can better differentiate the occlusion on the leg when compared to the other locations foe which the traditional ABI is in the normal range.
Material and Method: In total 130 patients suspected to have PAOD were recruited and they underwent Doppler ultrasound method to diagnose PAOD and the traditional ABI was calculated using the pressure cuff. Then the PPG signals were recorded simultaneously from the 4 locations (right and left index fingers and toes) for 10seconds, and then four 10-second sequential PPG signals were also recorded from the same locations using a single probe. The noise of PPG signals was removed by a bandpass filter (1-10 Hz) and after peak detection, the peak of the pulsatile waveforms was aligned and averaged. The averaged waveform was used to extract the waveform onset, Peak, inflection point, and end point from the second derivative of the signal. Extracted multiple features from the PPG and used those for the classification. Found the PPG features exhibit AUC of 0.8481, 74% sensitivity, 84% Specificity whereas the traditional ABI has AUC of 0.7914, 52% sensitivity, 98% Specificity. Found the PPG features have better Sensitivity and AUC when compared to the traditional ABI. Hence to improve the results extracted more features from the signal and did the classification. Found the PPG features have 0.8448 AUC, Sensitivity of 73%, specificity of 84% whereas the traditional ABI has 0.7692 AUC, Sensitivity of 48%, Specificity of 98% at the optimum threshold. The classification of the arterial occlusion location with the PPG features gave better results when compared to the traditional ABI. PPG features can identify 91.6% of patients above knee and found 65% of the patients below knee whereas at the optimum threshold of 0.9 the ABI found 75% of patients above knee and 36.7% of the patients below knee. Then Kurtosis and the Diastolic with at the 90% of the pulse height gave the better differentiation for the arteries on the above and below knee region.
Discussion: The PPG features exhibit better Sensitivity and AUC is more when compared to the traditional ABI. But the Specificity is less of the PPG features when compared to the traditional ABI. One advantage of this system is here we have used the PPG signal from the toes compared to the previous studies and done the classification using the Ultrasound results which is more reliable other than comparing it with the range of the ABI values. With the minimum amount of data on the above region the Kurtosis gave the better differentiation at the threshold of 1.7208 and the Diastolic width provided the better results for the arteries below knee. Because the nearly 30% of the PPG signal obtained from the Patients with PAOD are noisy due to the severity of the occlusion and not able to get any useful information from those values. Since the PPG signal cannot be acquired when the Occlusion is more severe. Hence it can be used as tool for the initial investigation of the PAOD detection where there is no need of complex measuring techniques are required. When we classify the patients according to the location of the occlusion such as the arteries above and below knee at the optimum threshold found for the PPG features gave better identification results when compared to the traditional ABI. Hence PPG can be used as an initial investigation tool for the identification of occlusion in all arterial position gave promising results when compared to the traditional ABI.
Incorporating deep learning techniques could offer potential improvements, paving the way for more accurate and efficient PAOD diagnosis and treatment planning, ultimately benefiting patient outcomes and healthcare practices. |
