An integrative statistics-guided image-based multi-scale lung model

描述（由申请人提供）：本研究的最终目标是通过一个多尺度计算流体动力学（CFD）肺模型，将群体数据的统计分析与个体受试者的功能预测相结合，建立一个新的肺功能评估和预测计算框架，以改善患者表型，从而实现患者特异性治疗。推动这项研究的一个假设是，肺表型可能因性别、年龄和（正常或患病）状态而表现出相似的特征，因此它们可以聚集成亚群，亚群的结构和功能特征可能与吸入颗粒物的沉积和肺部炎症有关。为了实现目标和检验假设，我们提出以下具体目标。(1)对气道图像测量及相关协变量进行统计分析。(2)进行图像配准分析，研究区域通气、组织分数和肺变形。(3)针对病变肺开发多尺度受试者特异性气道树建模和网格算法。(4)采用并行CFD模型研究气道阻力、颗粒沉积和热点。热点是吸入颗粒、毒素、刺激物或细菌在肺部积聚的地方。(5)从人体研究中寻求支持性数据，以证明CFD模型可以预测与吸入颗粒物沉积增强相关的肺部炎症易感区域。我们建议分析现有的和正在增长的庞大数据库，如肺部计算机断层扫描（CT）图像数据、人口统计信息、吸烟史和肺功能测试，这些数据都是由NIH资助的多中心试验收集的。统计方法将

DESCRIPTION (provided by applicant): The ultimate goal of the research is to build a new computational framework for assessment and prediction of lung function through integration of statistical analysis of population data with prediction of function in individual subjects via a muti-scale computational fluid dynamics (CFD) lung model, for improved patient phenotyping and hence patient-specific therapy. An hypothesis motivating this research is that lung phenotypes may exhibit similar features by gender, age, and (normal or diseased) state, thus they can be clustered into sub- populations, and the structural and functional features in sub-populations may correlate with deposition of inhaled particulates and inflammation in the lungs. To achieve the goal and test the hypothesis, we propose the following specific aims. (1) Perform statistical analysis of airway image-based measurements and associated covariates. (2) Perform image registration analysis to study regional ventilation, tissue fraction and lung deformation. (3) Develop multi-scale subject-specific airway tree modeling and meshing algorithms for diseased lungs. (4) Apply a parallel CFD model to study airway resistance, particle deposition, and hot spots. Hot spots are the locations where inhaled particles, toxins, irritants, or bacteria accumulate in the lungs. (5) Seek supportive data from human studies to demonstrate that CFD modeling predicts lung regions susceptible to inflammation associated with enhanced deposition of inhaled particulate. We propose to analyze the existing and growing huge databases, such as lung computed tomography (CT) image data, demographic information, smoking history, and pulmonary function tests, collected by the NIH funded multi-center trials. Statistical methods will be applied to cluster and classify large data sets into sub-populations. The novelty of our approach lies in fusion of both static structural and dynamic functional phenotypes into our statistical analyses, including morphologic and topological airway measurements and threshold-based measurements of air trapping and emphysema extracted from a single CT lung image, deformation-based functional variables derived from image registration of CT images at two lung volumes, and CFD-predicted sensitive functional variables. These statistical tools will identify statistically significant phenotypes contrasting normal, COPD and asthmatic subjects, and identify a few subjects representative of sub-populations for multi-scale high- performance parallel CFD simulations to study flows, resistance, and hot spots, and their correlations with the inflammations of airways and tissues. Human subject studies will be conducted using volumetric 3D lung dual energy computed tomography (DECT) and 99mTc-MPAO-labelled white blood cell (WBC) lung SPECT imaging for model validation and longitudinal studies.