School of Engineering, Brown University, Providence, RI, 02912, USA.
Center for Biomedical Engineering, Brown University, Providence, RI, 02912, USA.
Sci Rep. 2018 Apr 3;8(1):5581. doi: 10.1038/s41598-018-23488-y.
Spatiotemporal tracking of tracer particles or objects of interest can reveal localized behaviors in biological and physical systems. However, existing tracking algorithms are most effective for relatively low numbers of particles that undergo displacements smaller than their typical interparticle separation distance. Here, we demonstrate a single particle tracking algorithm to reconstruct large complex motion fields with large particle numbers, orders of magnitude larger than previously tractably resolvable, thus opening the door for attaining very high Nyquist spatial frequency motion recovery in the images. Our key innovations are feature vectors that encode nearest neighbor positions, a rigorous outlier removal scheme, and an iterative deformation warping scheme. We test this technique for its accuracy and computational efficacy using synthetically and experimentally generated 3D particle images, including non-affine deformation fields in soft materials, complex fluid flows, and cell-generated deformations. We augment this algorithm with additional particle information (e.g., color, size, or shape) to further enhance tracking accuracy for high gradient and large displacement fields. These applications demonstrate that this versatile technique can rapidly track unprecedented numbers of particles to resolve large and complex motion fields in 2D and 3D images, particularly when spatial correlations exist.
示踪粒子或感兴趣物体的时空跟踪可以揭示生物和物理系统中的局部行为。然而，现有的跟踪算法对于经历的位移小于其典型粒子间分离距离的相对较少的粒子最为有效。在这里，我们展示了一种单粒子跟踪算法，可以重建具有大量粒子的大复杂运动场，粒子数量比以前可解析的数量级大得多，从而为在图像中实现非常高的奈奎斯特空间频率运动恢复打开了大门。我们的关键创新是编码最近邻位置的特征向量、严格的异常值去除方案和迭代变形变形方案。我们使用合成和实验生成的 3D 粒子图像测试了该技术的准确性和计算功效，包括软材料中的非仿射变形场、复杂的流体流动和细胞产生的变形。我们通过添加额外的粒子信息（例如颜色、大小或形状）来增强该算法，以进一步提高高梯度和大位移场的跟踪精度。这些应用表明，这种通用技术可以快速跟踪前所未有的大量粒子，以解决二维和三维图像中大而复杂的运动场问题，特别是在存在空间相关性时。
