Analysis of cellular objects through diffraction images acquired by flow cytometry

In situ monitoring of toxic effects of algal toxin on cells by a novel microfluidic flow cytometry instrument

Algal toxins produced by microalgae, such as domoic acid (DA)1, have toxic effects on humans. However, toxicity tests using mice only yield lethal doses of algal toxins without providing insights into the mechanism of action on cells. In this study, a fast segmentation of microfluidic flow cytometry cell images based on the bidirectional background subtraction (BBS)2 method was developed to get the visual evidence of apoptosis in both bright-field and fluorescence images. This approach enables mapping of changes in cell morphology and activity under algal toxins, allowing for fast (within 60 s) and automated biological detection. By combining microfluidics with flow cytometry, the intricate cellular-level reaction process can be observed in micro samples of 293 T cells and mouse spleen cells, offering potential for future in vitro experiments.

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

Development of label-free methods for accurate classification of cells with high throughput can yield powerful tools for biological research and clinical applications. We have developed a deep neural network of DINet for extracting features from cross-polarized diffraction image (p-DI) pairs on multiple pixel scales to accurately classify cells in five types. A total of 6185 cells were measured by a polarization diffraction imaging flow cytometry (p-DIFC) method followed by cell classification with DINet on p-DI data. The averaged value and SD of classification accuracy were found to be 98.9% ± 1.00% on test data sets for 5-fold training and test. The invariance of DINet to image translation, rotation, and blurring has been verified with an expanded p-DI data set. To study feature-based classification by DINet, two sets of correctly and incorrectly classified cells were selected and compared for each of two prostate cell types. It has been found that the signature features of large dissimilarities between p-DI data of correctly and incorrectly classified cell sets increase markedly from convolutional layers 1 and 2 to layers 3 and 4. These results clearly demonstrate the importance of high-order correlations extracted at the deep layers for accurate cell classification.

A Case Study of the Augmentation and Evaluation of Training Data for Deep Learning

Deep learning has been widely used for extracting values from big data. As many other machine learning algorithms, deep learning requires significant training data. Experiments have shown both the volume and the quality of training data can significantly impact the effectiveness of the value extraction. In some cases, the volume of training data is not sufficiently large for effectively training a deep learning model. In other cases, the quality of training data is not high enough to achieve the optimal performance. Many approaches have been proposed for augmenting training data to mitigate the deficiency. However, whether the augmented data are “fit for purpose” of deep learning is still a question. A framework for comprehensively evaluating the effectiveness of the augmented data for deep learning is still not available. In this article, we first discuss a data augmentation approach for deep learning. The approach includes two components: the first one is to remove noisy data in a dataset using a machine learning based classification to improve its quality, and the second one is to increase the volume of the dataset for effectively training a deep learning model. To evaluate the quality of the augmented data in fidelity, variety, and veracity, a data quality evaluation framework is proposed. We demonstrated the effectiveness of the data augmentation approach and the data quality evaluation framework through studying an automated classification of biology cell images using deep learning. The experimental results clearly demonstrated the impact of the volume and quality of training data to the performance of deep learning and the importance of the data quality evaluation. The data augmentation approach and the data quality evaluation framework can be straightforwardly adapted for deep learning study in other domains.

Light-sheet-based 2D light scattering cytometry for label-free characterization of senescent cells

A light-sheet-based 2D light scattering cytometer is developed for label-free characterization of senescent cells. The light-sheet provides an illumination beam with controlled thickness for single cell excitation, and 2D light scattering patterns are obtained by using a defocused imaging method. The principle of this cytometer is validated by distinguishing microspheres with submicron resolution. Automatic classification of senescent and normal cells is achieved at single cell level by using the support vector machine (SVM) algorithm, where a sensitivity of 89.1% and a specificity of 96.4% are obtained. Our results suggest that the light-sheet-based 2D light scattering label-free cytometry has the capability to perform size differentiation of beads with submicron resolution and to classify different groups of cells without fluorescent labeling, showing the potential for clinical diagnosis of senescence-related diseases.

Differentiation of normal and leukemic cells by 2D light scattering label-free static cytometry

Two-dimensional (2D) light scattering patterns of single microspheres, normal granulocytes and leukemic cells are obtained by label-free static cytometry. Statistical results of experimental 2D light scattering patterns obtained from standard microspheres with a mean diameter of 4.19 μm agree well with theoretical simulations. High accuracy rates (greater than 92%) for label-free differentiation of normal granulocytes and leukemic cells, both the acute and chronic leukemic cells, are achieved by analyzing the 2D light scattering patterns. Our label-free static cytometry is promising for leukemia screening in clinics.

Comparison study of distinguishing cancerous and normal prostate epithelial cells by confocal and polarization diffraction imaging

Accurate classification of malignant cells from benign ones can significantly enhance cancer diagnosis and prognosis by detection of circulating tumor cells (CTCs). We have investigated two approaches of quantitative morphology and polarization diffraction imaging on two prostate cell types to evaluate their feasibility as single-cell assay methods toward CTC detection after cell enrichment. The two cell types have been measured by a confocal imaging method to obtain their three-dimensional morphology parameters and by a polarization diffraction imaging flow cytometry (p-DIFC) method to obtain image texture parameters. The support vector machine algorithm was applied to examine the accuracy of cell classification with the morphology and diffraction image parameters. Despite larger mean values of cell and nuclear sizes of the cancerous prostate cells than the normal ones, it has been shown that the morphologic parameters cannot serve as effective classifiers. In contrast, accurate classification of the two prostate cell types can be achieved with high classification accuracies on measured data acquired separately in three measurements. These results provide strong evidence that the p-DIFC method has the potential to yield morphology-related “fingerprints” for accurate and label-free classification of the two prostate cell types.

Original Research: Label-free detection for radiation-induced apoptosis in glioblastoma cells

Current flow cytometry (FCM) requires fluorescent dyes labeling cells which make the procedure costly and time consuming. This manuscript reports a feasibility study of detecting the cell apoptosis with a label-free method in glioblastoma cells. A human glioma cell line M059K was exposed to 8 Gy dose of radiation, which enables the cells to undergo radiation-induced apoptosis. The rates of apoptosis were studied at different time points post-irradiation with two different methods: FCM in combination with Annexin V-FITC/PI staining and a newly developed technique named polarization diffraction imaging flow cytometry. Totally 1000 diffraction images were acquired for each sample and the gray level co-occurrence matrix (GLCM) algorithm was used in morphological characterization of the apoptotic cells. Among the feature parameters extracted from each image pair, we found that the two GLCM parameters of angular second moment (ASM) and sum entropy (SumEnt) exhibit high sensitivities and consistencies as the apoptotic rates (Pa) measured with FCM method. In addition, no significant difference exists between Pa and ASM_S, Pa and SumEnt_S, respectively (P > 0.05). These results demonstrated that the new label-free method can detect cell apoptosis effectively. Cells can be directly used in the subsequent biochemical experiments as the structure and function of cells and biomolecules are well-preserved with this new method.

Realistic optical cell modeling and diffraction imaging simulation for study of optical and morphological parameters of nucleus

Coherent light scattering presents complex spatial patterns that depend on morphological and molecular features of biological cells. We present a numerical approach to establish realistic optical cell models for generating virtual cells and accurate simulation of diffraction images that are comparable to measured data of prostate cells. With a contourlet transform algorithm, it has been shown that the simulated images and extracted parameters can be used to distinguish virtual cells of different nuclear volumes and refractive indices against the orientation variation. These results demonstrate significance of the new approach for development of rapid cell assay methods through diffraction imaging.

Acquisition of cross-polarized diffraction images and study of blurring effect by one time-delay-integration camera

Blurred diffraction images acquired from flowing particles affect the measurement of fringe patterns and subsequent analysis. An imaging unit with one time-delay-integration (TDI) camera has been developed to acquire two cross-polarized diffraction images. It was shown that selected elements of Mueller matrix of single scatters can be imaged with pixel matching precision in this configuration. With the TDI camera, the effect of blurring on imaging of scattered light propagating along the side directions was found to be much more significant for biological cells than microspheres. Despite blurring, classification of MCF-7 and K562 cells is feasible since the effect has similar influence on extracted image parameters. Furthermore, image blurring can be useful for analysis of the correlations among texture parameters for characterization of diffraction images from single cells. The results demonstrate that with one TDI camera the polarization diffraction imaging flow cytometry can be significantly improved and angular distribution of selected Mueller matrix elements can be accurately measured for rapid and morphology-based assay of particles and cells without fluorescent labeling.

Analysis of diffraction imaging in non-conjugate configurations

Diffraction imaging of scattered light allows extraction of information on scatterer's morphology. We present a method for accurate simulation of diffraction imaging of single particles by combining rigorous light scattering model with ray-tracing software. The new method has been validated by comparison to measured images of single microspheres. Dependence of fringe patterns on translation of an objective based imager to off-focus positions has been analyzed to clearly understand diffraction imaging with multiple optical elements. The calculated and measured results establish unambiguously that diffraction imaging should be pursued in non-conjugate configurations to ensure accurate sampling of coherent light distribution from the scatterer.

Polarization imaging and classification of Jurkat T and Ramos B cells using a flow cytometer

Label-free and rapid classification of cells can have awide range of applications in biology. We report a robust method of polarization diffraction imaging flow cytometry (p-DIFC) for achieving this goal. Coherently scattered light signals are acquired from single cells excited by a polarized laser beam in the form of two cross-polarized diffraction images. Image texture and intensity parameters are extracted with a gray level co-occurrence matrix (GLCM) algorithm to obtain an optimized set of feature parameters as the morphological "fingerprints" for automated cell classification. We selected the Jurkat T cells and Ramos B cells to test the p-DIFC method's capacity for cell classification. After detailed statistical analysis, we found that the optimized feature vectors yield accuracies of classification between the Jurkat and Ramos ranging from 97.8% to 100% among different cell data sets. Confocal imaging and three-dimensional reconstruction were applied to gain insights on the ability of p-DIFC method for classifying the two cell lines of highly similar morphology. Based on these results we conclude that the p-DIFC method has the capacity to discriminate cells of high similarity in their morphology with "fingerprints" features extracted from the diffraction images, which may be attributed to subtle but statistically significant differences in the nucleus-to-cell volume ratio in the case of Jurkat and Ramos cells.

A quantitative method for measurement of HL-60 cell apoptosis based on diffraction imaging flow cytometry technique

A quantitative method for measurement of apoptosis in HL-60 cells based on polarization diffraction imaging flow cytometry technique is presented in this paper. Through comparative study with existing methods and the analysis of diffraction images by a gray level co-occurrence matrix algorithm (GLCM), we found 4 GLCM parameters of contrast (CON), cluster shade (CLS), correlation (COR) and dissimilarity (DIS) exhibit high sensitivities as the apoptotic rates. It was further demonstrated that the CLS parameter correlates significantly (R(2) = 0.899) with the degree of nuclear fragmentation and other three parameters showed a very good correlations (R(2) ranges from 0.69 to 0.90). These results demonstrated that the new method has the capability for rapid and accurate extraction of morphological features to quantify cellular apoptosis without the need for cell staining.