3-D Imaging Flow Cytometry

点击翻译按钮获取中文摘要

PROJECT SUMMARY This project aims to develop and test two innovative platforms and related software for 3-D imaging flow cytometry of fluorescent or absorbing (stained) samples. These systems will allow 3-D structural and functional imaging of many single cells at a subcellular resolution and at a scale that used to be available only in flow cytometry or recently in 2-D imaging. Thereby, the proposed methods have the potential to fundamentally change the ways cultured cells, patient-derived samples, and small experimental organisms are studied. Automated classification based on the 3-D features will enable the diagnosis of hematologic disorders at single-cell precision. Existing 3-D microscopy methods can provide the same information at higher resolution; however, by relying on a scanning mechanism they cannot be applied to suspending cells, especially in a flow configuration, which is essential for high-speed interrogation. Snapshot 3-D microscopy techniques have been developed to address this challenge, but they have insufficient spatial resolution for single-cell imaging and suffer from long data processing time. We overcome these limitations by combining two novel snapshot techniques developed by the PI with the most rigorous optical imaging theories and cutting-edge component technologies. We will use an array of lenslets, which simultaneously records many projection images corresponding with different viewing angles. The use of pupil phase masks, designed using wavefront coding and a theory of 3-D high-numerical-aperture optical imaging, will increase the resolution of each projection image to the theoretical limit given by the objective-lens numerical aperture. The target resolution is 0.5 µm, which is comparable to existing 2-D imaging flow cytometry systems. The target imaging throughputs based on current component technologies are 120 volumes/sec for fluorescence imaging and 700 volumes/sec for absorption imaging, which are higher than 100 volumes/sec of cutting-edge 3-D optical microscopy for stationary specimens. The vast amount of data acquired by these 3-D imaging systems imposes a serious challenge to data processing. The developed systems record true projection images, which obviate iterative deconvolution process, thereby allowing much faster tomographic reconstruction than in existing snapshot techniques. Using general-purpose graphics processing units and optical diffraction tomography, which includes the diffraction of light by subcellular organelles, our tomographic reconstruction algorithm will be faster yet more accurate than existing approaches. Further, we will explore the feasibility of applying a deep convolutional neural network to the images acquired by the developed systems for accurate single-cell classification based on 3-D features.