4D Shape Analysis for Modeling Spatiotemporal Change Trajectories in Huntington's

描述(申请人提供)：亨廷顿病(HD)是一种遗传性神经退行性疾病。主要研究结果描述了神经退化的进行性和临床前阶段非常早的细微变化，例如，神经纹状体和其他皮质下结构的萎缩，以及在临床症状被诊断之前长达15年的运动异常观察。为了更好地理解先兆阶段潜在的神经生物学和认知变化的时间进程，这对开发新的治疗方法至关重要，促使了多中心预测-HD研究的纵向设计。有HD风险的受试者被反复成像和检查，以研究患者特定的大脑结构轨迹和相关的认知变化。鉴于该联盟拥有庞大的纵向成像数据数据库，显然有必要制定敏感的措施来描述和描述这种变化的时机和性质。这项在Predicate-HD中进行辅助研究的建议将提供新开发的计算解剖学工具，专门用于分析纵向图像数据中的解剖结构，并用于对大脑形态测量的时空轨迹进行统计建模。我们将通过与Predicate-HD联盟共享的计算环境，为纵向神经影像数据提供时空(4D)分析的新工具。所提出的方法在分析和计算方面具有特别的创新性：(1)通过连续建模克服了纵向研究的局限性，克服了多个非均匀分布的时间点和丢失数据的固有挑战；(2)提供了高效和健壮的4D形状建模，而不需要跨形状组计算对应的地标；(3)应用模拟生物生长的数学概念来保证平滑的4D形状轨迹；以及(4)在感兴趣的结构嵌入其解剖环境中的多对象复杂的联合分析。这一新资源将显著提高Predicate-HD联盟的图像分析能力，因为这些工具将提供影响单个或多个对象皮质下结构的病理生理过程的时间进程的建模，为研究人员提供对病理时间进程和进展的新见解。该项目将在爱荷华州的MRI分割工作、我们新颖的4D形状建模方法以及犹他州在统计形状分析和时空形状建模方面的专业知识之间提供最佳合作，并结合爱荷华州Predicate-HD联盟在纵向数据分析方面的生物统计学优势。这种协同作用包括两个小组在提供可共享的计算资源和培训材料方面的强大专业知识。除了提供工具外，我们的合作还将处理和分析拥有多达351个多时间点MRI数据集的大型预测-高清数据库。这可能会导致新的生物标记物，这些生物标记物对 开发新的治疗方法来预防或减缓症状的进展。这一资源也服务于普通科学界，因为它是通用的W.r.t.应用程序领域，并通过NITRC免费分发。 公共卫生相关性：亨廷顿病(HD)是一种遗传性疾病，在北美有3万人受到影响，15万人面临患病风险。之前的神经成像研究表明，进行性脑萎缩在症状严重到足以确保可靠诊断的许多年前就开始了。鉴于开发新的药物疗法可以延缓甚至预防疾病的发生或减缓疾病的进展，开发这种变化的敏感的客观生物标记物尤为重要。使用来自多点预测-HD联盟的纵向成像数据，我们提出了新的创新的时空形状分析方法，该方法提供了个体大脑变化的时间进程的详细表征。工具包和培训材料将为Forecast-HD联盟以及科学界提供新的资源。

DESCRIPTION (provided by applicant): Huntington's Disease (HD) is an inherited, neurodegenerative disorder. Major research findings describe the progressive nature of neurodegeneration and subtle changes very early during the preclinical phase, e.g. atrophy of the neurostriatum and other subcortical structures and observation of movement abnormalities up to 15 years before clinical symptoms are diagnosed. The need for improved understanding of the time-course of underlying neurobiological and cognitive changes during the prodromal stage, which is essential for the development of new therapies, motivated the longitudinal design of the multi-center PREDICT-HD study. Subjects at risk for HD are imaged and examined repeatedly to study patient-specific trajectories of brain structures and associated cognitive changes. Given the consortium's large database of longitudinal imaging data, there is a clear need to develop sensitive measures describing and characterizing the timing and nature of such changes. This proposal for an Ancillary Study in PREDICT-HD will provide newly developed computational anatomy tools specifically designed for the analysis of anatomical structures in longitudinal image data and for the statistical modeling of spatio-temporal trajectories of morphometric brain measurements. We will provide novel tools for spatio-temporal (4D) analysis of longitudinal neuroimage data, via a shareable computational environment with the PREDICT-HD consortium. The proposed methods are particularly innovative in various analytical and computational aspects: (1) Overcoming limitations of longitudinal studies with its inherent challenges of multiple non-uniformly spaced time points and missing data via continuous modeling; (2) Presenting efficient and robust 4D shape modeling without the need to compute corresponding landmarks across shape-groups; (3) Applying a mathematical concept that mimics biological growth to guarantee smooth 4D shape trajectories, and (4) The joint analysis of multi-object complexes where structures of interest are embedded in their anatomical context. This new resource will significantly enhance image-analysis capabilities of the PREDICT-HD consortium, as the tools will provide a modeling of the time course of pathophysiological processes affecting single or multi-object subcortical structures, offering researchers new insight into the time-course and progression of pathology. This project will provide optimal collaboration between MRI segmentation work at Iowa, our novel 4D shape modeling methodology as well as the expertise on statistical shape analysis and spatiotemporal shape modeling at Utah, combined with biostatistical excellence in longitudinal data analysis of the PREDICT-HD consortium at Iowa. This synergy includes both groups' strong expertise in providing shareable computational resources and training materials. Beyond providing tools, our collaborative efforts will process and analyze the large PREDICT-HD data- base, with up to 351 multi-time point MRI datasets. This will potentially lead to new biomarkers that are crucial to the development of new therapies to prevent onset or slow the progression of symptoms. This resource also serves the general scientific community since it is generic w.r.t. the application domain and freely distributed via NITRC. PUBLIC HEALTH RELEVANCE: Huntington's disease (HD) is a genetic, hereditary disease with 30,000 people in North America affected and150,000 at risk for illness. Previous neuroimaging research showed progressive brain atrophy that begins many years before symptoms are severe enough to ensure reliable diagnosis. In view of developing new drug therapies that may delay or even prevent disease onset or slow down disease progression, it is particularly important to develop sensitive objective biomarkers of such changes. Using longitudinal imaging data from the multi-site PREDICT-HD consortium, we propose new innovative spatio-temporal shape analysis methodology that provides a detailed characterization of the time course of brain changes of individuals. The toolkit and training materials will represent a new resource for the PREDICT-HD consortium but also for the scientific community.