Reconstruction From Multiple Poses in Fluorescence Imaging: Proof of Concept
Reconstruction From Multiple Poses in Fluorescence Imaging: Proof of Concept
复制标题
DOI：
10.1109/jstsp.2015.2493884
复制
发表时间：
2016-02-01
影响因子：
7.5
通讯作者：
Unser, Michael
中科院分区：
文献类型：
--
作者：
Fortun, Denis;Guichard, Paul;Unser, Michael
所有荧光成像方式的一个缺点是轴向分辨率较差。为了解决这个问题，我们提出了一种新的方法来重建各向同性的高分辨率的荧光体积，从多个方向观察到的粒子复制的图像。我们设计了一种用于三维荧光成像的联合反卷积和多视角重建方法。我们通过设计一个快速的增广拉格朗日优化器来解决与大数据相关的计算挑战。该算法迭代部分的计算代价不依赖于输入粒子的数量。我们通过实验证明了我们的框架带来的分辨率提高，以及它处理实际约束的能力，比如大的PSF尺寸和大量的粒子。验证是在真实的模拟数据上执行的，这为我们的框架建立了概念证明，并将其定义为未来扩展的基础。
One disadvantage of all fluorescence imaging modalities is a poor axial resolution. To overcome this issue, we propose a novel approach to reconstruct fluorescence volumes with isotropic high resolution, from images of particle replicates observed at multiple orientations. We design a joint deconvolution-and-multiview reconstruction approach dedicated to three-dimensional fluorescence imaging. We address the computational challenge associated to big data by designing a fast augmented-Lagrangian optimizer. The computational cost of the iterative part of the algorithm does not depend on the number of input particles. We experimentally demonstrate the resolution improvement yielded by our framework and its ability to handle practical constraints like large PSF sizes and large number of particles. The validation is performed on realistic simulated data, which establishes a proof of concept for our framework and defines it as the basis for future extensions.