電腦視覺與血流動力學結合應用於冠狀動脈狹窄之定量評估

Abstract

冠狀動脈狹窄是由於動脈粥狀硬化所導致的疾病，臨床上會根據血管狹窄的程度判斷是否需執行介入性手術，但目前大部分的診斷是以外觀性的醫學影像作為觀察的依據，由於缺乏生理功能性的資訊，因此難以進行準確的評估。血流儲備分數為現今診斷心肌缺血功能的黃金指標，在最大血流的情況下，透過可量測壓力的導絲進入到冠狀動脈的病灶處，量測病灶遠端相對於近端正常血管平均壓力的比值，藉此數值描述病灶處對血管功能影響的程度，具有高度的準確性，但由於手術過程繁複且執行時須施打藥劑，因此難以順利推廣。近年來，有研究提出以三維重建血管模型與血流動力學建構的定量血流比率檢測技術，該方法不需使用壓力導絲即可透過模擬推算出虛擬的血流儲備分數，具有與血流儲備分數匹敵的準確度，且同時降低手術所需的成本、風險以及時間，對於非侵入式的功能性評估具有很高的潛力以及發展空間。 本研究建立半自動化檢測冠狀動脈定量血流比率的系統，透過圖形使用者介面手動選取血管攝影影格，並藉由整合影像分割技術提取血管的結構，接著選取感興趣的血管區段，利用投影的方式自動重建出相對應的三維網格立體模型，同時輸出病灶的外觀資訊，最後將血管模型匯入至計算流體力學模擬軟體進行血流動力學的分析，藉此計算出血管病灶處遠端相對於近端的壓力比值，以非侵入性的方式進行生理功能性的評估。 本研究使用實際進行過血管攝影的病患資料作為測試樣本，並針對模擬結果作初步的檢測。結果顯示，計算出的模擬值均符合臨床資料的主要趨勢，且系統整體的運算時間均可於4分鐘內完成，符合臨床效率的需求，可望提供醫療人員診斷冠狀動脈狹窄的輔助工具，並給予適當的治療方式。

Coronary artery stenosis is a disease caused by arteriosclerosis. In clinical practice, interventional surgery is performed according to the degree of vascular stenosis. However, most of the current diagnosis is based on angiographic appearance. Due to the lack of physiological and functional information, it is difficult to make an accurate assessment. Fractional flow reserve (FFR) is the gold standard for the diagnosis of myocardial ischemia. It is done through a pressure guidewire delivered to the lesion of the coronary artery to measure the ratio of mean distal coronary pressure to mean aortic pressure under hyperemic flow. FFR is a highly reliable method in assessing the functional significance of coronary stenosis. However, due to the complicated procedure and requirement of pharmacological maximal vasodilation, it is difficult to promote the use of FFR. In recent years, quantitative flow ratio (QFR) based on 3-dimensional vessel reconstructions and hemodynamic estimates has been proposed as a novel method for physiological lesion assessment. QFR is a wire-free method for computation of virtual FFR. It shows good diagnostic accuracy compared with FFR, and reduces the costs, risks and time required by surgery. There is great potential of development for using QFR as a non-invasive functional assessment of coronary lesions. In this research, a semi-automatic system was established to detect the QFR of coronary arteries. Firstly, the frame of angiography was manually selected through the graphical user interface, and the image segmentation was applied to extract the vascular structure. Secondly, after selecting the region of interest of coronary artery, the corresponding 3-dimentional mesh model would be reconstructed automatically by projection and the lesion data were exported at the same time. Finally, the blood vessel model was imported to the computational fluid dynamics simulation software for hemodynamic analysis, so as to calculate the pressure ratio of the distal coronary pressure to the proximal coronary pressure, and evaluated the physiological function of lesion in a non-invasive way. In this research, patients who undergone coronary angiography were included as sample tests, and the simulation results were given a preliminary inspection. The results show that the calculated simulation values are in line with the main trend of clinical data, and the overall operation time of the system could be completed within 4 minutes, which significantly reduces the time and costs of surgery, and provides physicians with auxiliary tools for diagnosing coronary artery stenosis and appropriate treatment methods.