三維幾何於先天性心臟病之計算分析：從手術規劃、型態量化到血流動力學評估

Abstract

複雜性先天性心臟病具有極端複雜的解剖異質性，且大多數需要多次手術介入處理。由於解剖之變異性大，手術前之詳細規劃與血流動力學之預測格外重要，然而傳統二維影像難以完整呈現其三維空間關係。本論文建立一套漸進式之計算工程平台，應用於多種先天性心臟病及縱膈腔疾病，從三維手術模擬進階至幾何型態計算學，最終延伸至計算血流動力學評估，以優化複雜先天性心臟病的手術決策與預後預測。 在手術模擬階段，本研究針對極度複雜之上下型右心室雙出口構造，透過虛擬手術預測傳統修補對於右心室出入口之壓迫，並驗證大動脈根部轉位術之可行性。在兒童微創心臟手術規劃上，開發了幾何射線演算法將手術室視野量化為可見度分數。15例臨床驗證顯示模擬器預測之主動脈管路插管位置分數是決定術中視野困難度之最顯著因子 (rho=-0.710, p=0.003)。 在幾何型態計算學階段，針對先天性氣管狹窄手術推導出斜切式滑動氣管整形術之幾何公式，5例應用顯示術後截面積增加327.2%，長度縮短僅15.8%，優於文獻報告之40-50%。此外針對右心房異構症合併功能性單一心室此一罕見疾病，提出心尖偏移指數(AEI)作為手術後長期存活預後指標，cox-LASSO迴歸中AEI為唯一獨立預測因子 (HR=0.088, p=0.002)，交叉驗證C指數達0.764。且高AEI組存活率78.6%，顯著優於低AEI組之23.1% (p=0.005)。 在血流動力學計算階段，對6例心外型全肺靜脈回流異常術後病患建立8個計算流體力學模型並計算量化。壁面剪應力變異係數可顯著區分狹窄與非狹窄模型(p=0.0098)，DBSCAN演算法則依照剪應力之高群聚區域定位高剪應力熱區，且與實際手術術中之發現高度一致。 本論文建立了從影像重建到功能性預測之計算架構，將複雜的三維解剖構造轉化為可量化的臨床指標，為複雜性先天性心臟病之精準計算手術提供系統性的工程基礎。

Complex congenital heart diseases (CHD) exhibit extreme anatomical heterogeneity and frequently require multiple surgical interventions. Due to this high anatomical variability, detailed preoperative planning and hemodynamic predictions are critical; however, traditional 2D imaging fails to adequately capture their complex 3D spatial relationships. This thesis establishes a progressive computational engineering platform applied to various CHDs and mediastinal diseases. Advancing from 3D surgical simulation to geometric morphometrics, and ultimately to computational hemodynamics, this platform aims to optimize surgical decision-making and prognostic predictions for complex CHD. In the surgical simulation phase, virtual surgery was utilized on the highly complex superior-inferior double outlet right ventricle to predict the compression of the right ventricular inflow and outflow tracts caused by traditional repair, successfully validating the feasibility of aortic root translocation. For pediatric minimally invasive cardiac surgery planning, a geometric ray-casting algorithm was developed to quantify the surgical field of view into visibility scores. Clinical validation across 15 cases revealed that the simulator-predicted visibility score for the aortic cannulation site was a significant determinant of intraoperative visual difficulty (rho = -0.710, p = 0.003). In the geometric morphometrics phase, geometric formulas for oblique slide tracheoplasty were derived for congenital tracheal stenosis. Application in 5 cases demonstrated a 327.2% increase in postoperative cross-sectional area with only a 15.8% reduction in length, significantly outperforming the 40-50% length reduction reported in the literature. Furthermore, for the rare condition of right atrial isomerism (RAI) with a functional single ventricle, the Apical Eccentricity Index (AEI) was proposed as a prognostic indicator for long-term postoperative survival. In Cox-LASSO regression analysis, AEI emerged as the sole independent predictor (HR = 0.088, p = 0.002) with a cross-validated C-index of 0.764. The survival rate in the high-AEI group (78.6%) was significantly superior to that of the low-AEI group (23.1%, p = 0.005). In the hemodynamic computation phase, 8 computational fluid dynamics (CFD) models were established and quantified for 6 patients following the repair of extracardiac total anomalous pulmonary venous connection (TAPVC). The coefficient of variation of wall shear stress (WSS) significantly differentiated stenotic from non-stenotic models (p = 0.0098). Additionally, the DBSCAN algorithm localized high WSS hotspots based on high-clustering regions, demonstrating high consistency with actual intraoperative findings. In conclusion, this thesis establishes a comprehensive computational framework ranging from image reconstruction to functional prediction. By translating complex 3D anatomical structures into quantifiable clinical indices, it provides a systematic engineering foundation for precision computational surgery in complex CHD.