Cameron B M, Robb R A
Biomedical Imaging Resource, Mayo Clinic College of Medicine, USA.
Stud Health Technol Inform. 2004;98:40-5.
Generating patient specific dynamic models is complicated by the complexity of the motion intrinsic and extrinsic to the anatomic structures being modeled. Using a physics-based sequentially deforming algorithm, an anatomically accurate dynamic four-dimensional model can be created from a sequence of 3-D volumetric time series data sets. While such algorithms may accurately track the cyclic non-linear motion of the heart, they generally fail to accurately track extrinsic structural and non-cyclic motion. To accurately model these motions, we have modified a physics-based deformation algorithm to use a meta-surface defining the temporal and spatial maxima of the anatomic structure as the base reference surface. A mass-spring physics-based deformable model, which can expand or shrink with the local intrinsic motion, is applied to the metasurface, deforming this base reference surface to the volumetric data at each time point. As the meta-surface encompasses the temporal maxima of the structure, any extrinsic motion is inherently encoded into the base reference surface and allows the computation of the time point surfaces to be performed in parallel. The resultant 4-D model can be interactively transformed and viewed from different angles, showing the spatial and temporal motion of the anatomic structure. Using texture maps and per-vertex coloring, additional data such as physiological and/or biomechanical variables (e.g., mapping electrical activation sequences onto contracting myocardial surfaces) can be associated with the dynamic model, producing a 5-D model. For acquisition systems that may capture only limited time series data (e.g., only images at end-diastole/end-systole or inhalation/exhalation), this algorithm can provide useful interpolated surfaces between the time points. Such models help minimize the number of time points required to usefully depict the motion of anatomic structures for quantitative assessment of regional dynamics.
生成针对特定患者的动态模型会因所建模解剖结构的内在和外在运动的复杂性而变得复杂。使用基于物理的顺序变形算法，可以从一系列三维体积时间序列数据集中创建解剖结构精确的动态四维模型。虽然此类算法可能能够准确跟踪心脏的周期性非线性运动，但它们通常无法准确跟踪外在结构和非周期性运动。为了准确模拟这些运动，我们修改了基于物理的变形算法，使用定义解剖结构的时间和空间最大值的元表面作为基础参考表面。将基于质量弹簧物理的可变形模型应用于元表面，该模型可以随着局部内在运动而扩展或收缩，从而在每个时间点将这个基础参考表面变形为体积数据。由于元表面包含了结构的时间最大值，任何外在运动都被固有地编码到基础参考表面中，并允许并行执行时间点表面的计算。所得的四维模型可以进行交互式变换并从不同角度查看，展示解剖结构的空间和时间运动。使用纹理映射和逐顶点着色，可以将诸如生理和/或生物力学变量等附加数据（例如，将电激活序列映射到收缩的心肌表面）与动态模型相关联，从而生成五维模型。对于可能只捕获有限时间序列数据的采集系统（例如，仅舒张末期/收缩末期或吸气/呼气时的图像），该算法可以在时间点之间提供有用的插值表面。此类模型有助于最小化有效描绘解剖结构运动以进行区域动力学定量评估所需的时间点数。
