Development and evaluation of realistic optical cell models for rapid and label‐free cell assay by diffraction imaging

Abstract

Methods for rapid and label-free cell assay are highly desired in life science. Single-shot diffraction imaging presents strong potentials to achieve this goal as evidenced by past experimental results using methods such as polarization diffraction imaging flow cytometry. We present here a platform of methods toward solving these problems and results of optical cell model (OCM) evaluations by calculations and analysis of cross-polarized diffraction image (p-DI) pairs. Four types of realistic OCMs have been developed with two prostate cell structures and adjustable refractive index (RI) parameters to investigate the effects of cell morphology and index distribution on calculated p-DI pairs. Image patterns have been characterized by a gray-level co-occurrence matrix (GLCM) algorithm and four GLCM parameters and linear depolarization ratio δ L have been selected to compare calculated against measured data of prostate cells. Our results show that the irregular shapes of and heterogeneity in RI distributions for organelles play significant roles in the spatial distribution of scattered light by cells in comparison to the average RI values and their differences among the organelles. Discrepancies in GLCM and δ L parameters between calculated and measured p-DI data provide useful insight for understanding light scattering by single cells and improving OCM.

Deep learning of diffraction image patterns for accurate classification of five cell types

“…It should be noted that the linear depolarization ratio δ L is given by the averaged 12‐bit pixel value ratio of p‐ and s‐polarized diffraction images. We have shown that δ L can be used for cell classification for its dependence on the types and distributions of molecular dipoles . For this study, however, we import only the combined DIs into a CNN classifier without δ L .…”

“…The input images were first processed by an algorithm for extraction of texture parameters followed by a classifier operating in the parameter space. Different algorithms have been explored for extracting image texture parameters, which include gray level co‐occurrence matrix (GLCM), short‐time Fourier transform, contourlet and Gabor transforms . Despite the variations in effectiveness for classification of cells in two types, the parameter based approach requires labor intensive assessment and validation.…”

Machine learning of diffraction image patterns for accurate classification of cells modeled with different nuclear sizes

“…Among these, imaging with coherent light scattered by cells stands out for strong signals and ability to profile internal structures by the 3D distribution of refractive index (RI) [6][7][8][9][10][11][12]. As we have discussed previously [13], 3D reconstruction of RI distribution consists of interferogram or diffraction image acquisition, error-prone phase unwrapping and computationally expensive tomographic reconstruction. Accomplishment of these steps is very challenging, which needs modeling nucleated cells with numerous and highly heterogeneous intracellular organelles of substantially irregular shapes.…”

Single‐Particle Characterization by Elastic Light Scattering

“…[ 78,95–97 ] However, there are rare examples of the throughput of up to 1000 particles per second. [ 98,99 ]…”

“…Due to high CCD exposure time, the stream velocity has to be significantly smaller than in other flow methods, leading to a low throughput. [ 103,104 ] The method has been reliably applied to particles of tens of μm in size, [ 50,51,98,105–108 ] however, characterization for particle sizes from 1 to 100 µm has also been demonstrated. [ 100,109 ] Other side‐scattering configurations include angular ranges: , , [ 110,111 ] , , [ 112 ] and , .…”