Interpretable unsupervised learning enables accurate clustering with high-throughput imaging flow cytometry.

A primary challenge of high-throughput imaging flow cytometry (IFC) is to analyze the vast amount of imaging data, especially in applications where ground truth labels are unavailable or hard to obtain. We present an unsupervised deep embedding algorithm, the Deep Convolutional Autoencoder-based Clustering (DCAEC) model, to cluster label-free IFC images without any prior knowledge of input labels. The DCAEC model first encodes the input images into the latent representations and then clusters based on the latent representations. Using the DCAEC model, we achieve a balanced accuracy of 91.9% for human white blood cell (WBC) clustering and 97.9% for WBC/leukemia clustering using the 3D IFC images and 3D DCAEC model. Above all, although no human recognizable features can separate the clusters of cells with protein localization, we demonstrate the fused DCAEC model can achieve a cluster balanced accuracy of 85.3% from the label-free 2D transmission and 3D side scattering images. To reveal how the neural network recognizes features beyond human ability, we use the gradient-weighted class activation mapping method to discover the cluster-specific visual patterns automatically. Evaluation results show that the automatically identified salient image regions have strong cluster-specific visual patterns for different clusters, which we believe is a stride for the interpretable neural network for cell analysis with high-throughput IFCs.

Label-free image-encoded microfluidic cell sorter with a scanning Bessel beam.

The microfluidic-based, label-free image-guided cell sorter offers a low-cost, high information content, and disposable solution that overcomes many limitations in conventional cell sorters. However, flow confinement for most microfluidic devices is generally only one-dimensional using sheath flow. As a result, the equilibrium distribution of cells spreads beyond the focal plane of commonly used Gaussian laser excitation beams, resulting in a large number of blurred images that hinder subsequent cell sorting based on cell image features. To address this issue, we present a Bessel-Gaussian beam image-guided cell sorter with an ultra-long depth of focus, enabling focused images of >85% of passing cells. This system features label-free sorting capabilities based on features extracted from the output temporal waveform of a photomultiplier tube (PMT) detector. For the sorting of polystyrene beads, SKNO1 leukemia cells, and Scenedesmus green algae, our results indicate a sorting purity of 97%, 97%, and 98%, respectively, showing that the temporal waveforms from the PMT outputs have strong correlations with cell image features. These correlations are also confirmed by off-line reconstructed cell images from a temporal-spatial transformation algorithm tailored to the scanning Bessel-Gaussian beam.

Roles of MAP kinases in the regulation of bone matrix gene expressions in human osteoblasts by oscillatory fluid flow.

We investigated the effects of oscillatory flow in regulating the gene expressions of type I collagen (COL1, the main component of human bone tissues) and osteopontin (OPN, the key gene for calcium deposition) in human osteoblast-like (MG-63) cells, and the roles of mitogen-activated protein kinases (MAPKs) in this regulation. The cells were subjected to oscillatory flow (0.5 +/- 4 dyn/cm(2)) or kept under static condition for various time periods (15 min, 30 min, 1 h, 2 h, 4 h, 8 h, and 16 h). Oscillatory flow caused significant up-regulations of both COL1 and OPN gene expressions over the 16 h of study, and a transient activation of MAPKs was starting at 15 min and declining to basal level in 2 h. The flow-induction of COL1 was blocked by an ERK inhibitor (PD98059) and reduced by a JNK inhibitor (SP600125), whereas that of OPN was abolished by PD98059. Analysis of the cis-elements in the COL1 and OPN promoters suggests the involvement of transacting factors Elk-1 and AP-1 in the transcription regulation. The ERK inhibitor (PD98059) blocked Elk-1 phosphorylation, as well as COL1 and OPN gene expression. The JNK inhibitor (SP600125) abolished c-jun phosphorylation and COL1 expression. These results suggest that the flow-induction of OPN was mediated through the ERK-Elk1-OPN pathway, and that COL1 was regulated by both the ERK-Elk1-COL1 and JNK-c-JUN-COL1 pathway.