Intelligent image-activated cell sorting 2.0

Exploiting machine learning for bestowing intelligence to microfluidics

Intelligent microfluidics is an emerging cross-discipline research area formed by combining microfluidics with machine learning. It uses the advantages of microfluidics, such as high throughput and controllability, and the powerful data processing capabilities of machine learning, resulting in improved systems in biotechnology and chemistry. Compared to traditional microfluidics using manual analysis methods, intelligent microfluidics needs less human intervention, and results in a more user-friendly experience with faster processing. There is a paucity of literature reviewing this burgeoning and highly promising cross-discipline. Therefore, we herein comprehensively and systematically summarize several aspects of microfluidic applications enabled by machine learning. We list the types of microfluidics used in intelligent microfluidic applications over the last five years, as well as the machine learning algorithms and the hardware used for training. We also present the most recent advances in key technologies, developments, challenges, and the emerging opportunities created by intelligent microfluidics.

Sheathless acoustic based flow cell sorter for enrichment of rare cells

Cell enrichment is a powerful tool in many kinds of cell research, especially in applications with low abundance cell types. In this work, we developed a microfluidic fluorescence activated cell sorting device that was able to perform on-demand, low loss cell detection, and sorting. The chip utilizes three-dimensional acoustic standing waves to position all cells in the same fluid velocity regime without sheath. When the cells pass through a laser interrogation region, the scattering and fluorescent signals are detected, translated and transported to software. The target cells are then identified by gating on the plots. Short bursts of standing acoustic waves are triggered by order from PC to sort target cells within predefined gating region. For very low abundance and rare labeled lymphocytes mixed with high concentration unlabeled white blood cells (WBCs), (1-100 labeled lymphocytes are diluted in 10⁶ WBCs in 1 ml volume fluid), the device is able to remove more than 98% WBCs and recover labeled lymphocytes with efficiency of 80%. We further demonstrated that this device worked with real clinical samples by successfully isolating fetal nucleated red blood cells (FNRBCs) in the blood samples from pregnant women with male fetus. The obtained cells were sequenced and the expressions of (sex determining region Y) SRY genes were tested to determine fetal cell proportion. In genetic analysis, the proportion of fetal cells in the final picked sample is up to 40.64%. With this ability, the device proposed could be valuable for biomedical applications involving fetal cells, circulating tumor cells, and stem cells.

Review: tomographic imaging flow cytometry

Within the last decades, conventional flow cytometry (FC) has evolved as a powerful measurement method in clinical diagnostics, biology, life sciences and healthcare. Imaging flow cytometry (IFC) extends the power of traditional FC by adding high resolution optical and spectroscopic information. However, the conventional IFC only provides a 2D projection of a 3D object. To overcome this limitation, tomographic imaging flow cytometry (tIFC) was developed to access 3D information about the target particles. The goal of tIFC is to visualize surfaces and internal structures in a holistic way. This review article gives an overview of the past and current developments in tIFC.

Label-free single-cell isolation enabled by microfluidic impact printing and real-time cellular recognition

Analysis of cellular components at the single-cell level is important to reveal cellular heterogeneity. However, current technologies to isolate individual cells are either label-based or have low performance. Here, we present a novel technique by integrating real-time cellular recognition and microfluidic impact printing (MIP) to isolate single cells with high efficiency and high throughput in a label-free manner. Specifically, morphological characteristics of polystyrene beads and cells, computed by an efficient image processing algorithm, are utilized as selection criteria to identify target objects. Subsequently, each detected single-cell object in the suspension is ejected from the microfluidic channel by impact force. It has been demonstrated that the single-cell isolating system has the ability to encapsulate polystyrene beads in droplets with an efficiency of 95%, while for HeLa cells, this has been experimentally measured as 90.3%. Single-cell droplet arrays are generated at a throughput of 2 Hz and 96.6% of the cells remain alive after isolation. This technology has significant potential in various emerging applications, including single-cell omics, tissue engineering, and cell-line development.

Investigation of viscoelastic focusing of particles and cells in a zigzag microchannel

Microfluidic particle focusing has been a vital prerequisite step in sample preparation for downstream particle separation, counting, detection, or analysis, and has attracted broad applications in biomedical and chemical areas. Besides all the active and passive focusing methods in Newtonian fluids, particle focusing in viscoelastic fluids has been attracting increasing interest because of its advantages induced by intrinsic fluid property. However, to achieve a well-defined focusing position, there is a need to extend channel lengths when focusing micrometer-sized or sub-microsized particles, which would result in the size increase of the microfluidic devices. This work investigated the sheathless viscoelastic focusing of particles and cells in a zigzag microfluidic channel. Benefit from the zigzag structure of the channel, the channel length and the footprint of the device can be reduced without sacrificing the focusing performance. In this work, the viscoelastic focusing, including the focusing of 10 μm polystyrene particles, 5 μm polystyrene particles, 5 μm magnetic particles, white blood cells (WBCs), red blood cells (RBCs), and cancer cells, were all demonstrated. Moreover, magnetophoretic separation of magnetic and nonmagnetic particles after viscoelastic pre-focusing was shown. This focusing technique has the potential to be used in a range of biomedical applications.

Three-dimensional imaging on a chip using optofluidics light-sheet fluorescence microscopy

Volumetric, sub-micron to micron level resolution imaging is necessary to assay phenotypes or characteristics at the sub-cellular/organelle scale. However, three-dimensional fluorescence imaging of cells is typically low throughput or compromises on the achievable resolution in space and time. Here, we capitalise on the flow control capabilities of microfluidics and combine it with microoptics to integrate light-sheet based imaging directly into a microfluidic chip. Our optofluidic system flows suspended cells through a sub-micrometer thick light-sheet formed using micro-optical components that are cast directly in polydimethylsiloxane (PDMS). This design ensures accurate alignment, drift-free operation, and easy integration with conventional microfluidics, while providing sufficient spatial resolution, optical sectioning and volumetric data acquisition. We demonstrate imaging rates of 120 ms per cell at sub-μm resolution, that allow extraction of complex cellular phenotypes, exemplified by imaging of cell clusters, receptor distribution, and the analysis of endosomal size changes.

Microscopic impedance cytometry for quantifying single cell shape

In this work, we investigated the ability of impedance flow cytometry to measure the shape of single cells/particles. We found that the impedance pulses triggered by micro-objects that are asymmetric in morphology show a tilting trend, and there is no such a tilting trend for symmetric ones. Therefore, we proposed a new metric, tilt index, to quantify the tilt level of the impedance pulses. Through simulation, we found that the value of tilt index tends to be zero for perfectly symmetrical objects, while the value is greater than zero for asymmetrical ones. Also, this metric was found to be independent on the trajectories (i.e., lateral, and z-direction shift) of the target micro-object. In experiments, we adopted a home-made lock-in amplifier and performed experiments on 10 μm polystyrene beads and Euglena gracilis (E. gracilis) cells with varying shapes. The experimental results coincided with the simulation results and demonstrated that the new metric (tilt index) enables the impedance cytometry to characterize the shape single cells/particles without microscopy or other optical setups.

Morphological Indicator for Directed Evolution of Euglena gracilis with a High Heavy Metal Removal Efficiency

In the past few decades, microalgae-based bioremediation methods for treating heavy metal (HM)-polluted wastewater have attracted much attention by virtue of their environment friendliness, cost efficiency, and sustainability. However, their HM removal efficiency is far from practical use. Directed evolution is expected to be effective for developing microalgae with a much higher HM removal efficiency, but there is no non-invasive or label-free indicator to identify them. Here, we present an intelligent cellular morphological indicator for identifying the HM removal efficiency of Euglena gracilis in a non-invasive and label-free manner. Specifically, we show a strong monotonic correlation (Spearman's ρ = -0.82, P = 2.1 × 10^-5) between a morphological meta-feature recognized via our machine learning algorithms and the Cu²⁺ removal efficiency of 19 E. gracilis clones. Our findings firmly suggest that the morphology of E. gracilis cells can serve as an effective HM removal efficiency indicator and hence have great potential, when combined with a high-throughput image-activated cell sorter, for directed-evolution-based development of E. gracilis with an extremely high HM removal efficiency for practical wastewater treatment worldwide.

Droplet flow cytometry for single-cell analysis

The interrogation of single cells has revolutionised biology and medicine by providing crucial unparalleled insights into cell-to-cell heterogeneity. Flow cytometry (including fluorescence-activated cell sorting) is one of the most versatile and high-throughput approaches for single-cell analysis by detecting multiple fluorescence parameters of individual cells in aqueous suspension as they flow past through a focus of excitation lasers. However, this approach relies on the expression of cell surface and intracellular biomarkers, which inevitably lacks spatial and temporal phenotypes and activities of cells, such as secreted proteins, extracellular metabolite production, and proliferation. Droplet microfluidics has recently emerged as a powerful tool for the encapsulation and manipulation of thousands to millions of individual cells within pico-litre microdroplets. Integrating flow cytometry with microdroplet architectures surrounded by aqueous solutions (e.g., water-in-oil-in-water (W/O/W) double emulsion and hydrogel droplets) opens avenues for new cellular assays linking cell phenotypes to genotypes at the single-cell level. In this review, we discuss the capabilities and applications of droplet flow cytometry (DFC). This unique technique uses standard commercially available flow cytometry instruments to characterise or select individual microdroplets containing single cells of interest. We explore current challenges associated with DFC and present our visions for future development.

Microfluidic formation of crystal-like structures

Crystal-like structures find application in several fields ranging from biomedical engineering to material science. For instance, droplet crystals are critical for high throughput assays and material synthesis, while particle crystals are important for particles and cell encapsulation, Drop-seq technologies, and single-cell analysis. Formation of crystal-like structures relies entirely on the possibility of manipulating with great accuracy the micrometer-size objects forming the crystal. In this context, microfluidic devices offer versatile tools for the precise manipulation of droplets and particles, thus enabling fabrication of crystal-like structures that form due to hydrodynamic interactions among droplets or particles. In this review, we aim at providing an holistic representation of crystal-like structure formation mediated by hydrodynamic interactions in microfluidic devices. We also discuss the physical origin of these hydrodynamic interactions and their relation to parameters such as device geometry, fluid properties, and flow conditions.

Analysis of human chromosomes by imaging flow cytometry

Chromosomal analysis is traditionally performed by karyotyping on metaphase spreads, or by fluorescent in situ hybridization (FISH) on interphase cells or metaphase spreads. Flow cytometry was introduced as a new method to analyze chromosomes number (ploidy) and structure (telomere length) in the 1970s with data interpretation largely based on fluorescence intensity. This technology has had little uptake for human cytogenetic applications primarily due to analytical challenges. The introduction of imaging flow cytometry, with the addition of digital images to standard multi-parametric flow cytometry quantitative tools, has added a new dimension. The ability to visualize the chromosomes and FISH signals overcomes the inherent difficulties when the data is restricted to fluorescence intensity. This field is now moving forward with methods being developed to assess chromosome number and structure in whole cells (normal and malignant) in suspension. A recent advance has been the inclusion of immunophenotyping such that antigen expression can be used to identify specific cells of interest for specific chromosomes and their abnormalities. This capability has been illustrated in blood cancers, such as chronic lymphocytic leukemia and plasma cell myeloma. The high sensitivity and specificity achievable highlights the potential imaging flow cytometry has for cytogenomic applications (i.e., diagnosis and disease monitoring). This review introduces and describes the development, current status, and applications of imaging flow cytometry for chromosomal analysis of human chromosomes.

The in vitro micronucleus assay using imaging flow cytometry and deep learning

The in vitro micronucleus (MN) assay is a well-established assay for quantification of DNA damage, and is required by regulatory bodies worldwide to screen chemicals for genetic toxicity. The MN assay is performed in two variations: scoring MN in cytokinesis-blocked binucleated cells or directly in unblocked mononucleated cells. Several methods have been developed to score the MN assay, including manual and automated microscopy, and conventional flow cytometry, each with advantages and limitations. Previously, we applied imaging flow cytometry (IFC) using the ImageStream® to develop a rapid and automated MN assay based on high throughput image capture and feature-based image analysis in the IDEAS® software. However, the analysis strategy required rigorous optimization across chemicals and cell lines. To overcome the complexity and rigidity of feature-based image analysis, in this study we used the Amnis® AI software to develop a deep-learning method based on convolutional neural networks to score IFC data in both the cytokinesis-blocked and unblocked versions of the MN assay. We show that the use of the Amnis AI software to score imagery acquired using the ImageStream® compares well to manual microscopy and outperforms IDEAS® feature-based analysis, facilitating full automation of the MN assay.

Rare bioparticle detection via deep metric learning

Recent deep neural networks have shown superb performance in analyzing bioimages for disease diagnosis and bioparticle classification. Conventional deep neural networks use simple classifiers such as SoftMax to obtain highly accurate results. However, they have limitations in many practical applications that require both low false alarm rate and high recovery rate, e.g., rare bioparticle detection, in which the representative image data is hard to collect, the training data is imbalanced, and the input images in inference time could be different from the training images. Deep metric learning offers a better generatability by using distance information to model the similarity of the images and learning function maps from image pixels to a latent space, playing a vital role in rare object detection. In this paper, we propose a robust model based on a deep metric neural network for rare bioparticle (Cryptosporidium or Giardia) detection in drinking water. Experimental results showed that the deep metric neural network achieved a high accuracy of 99.86% in classification, 98.89% in precision rate, 99.16% in recall rate and zero false alarm rate. The reported model empowers imaging flow cytometry with capabilities of biomedical diagnosis, environmental monitoring, and other biosensing applications.

User-friendly image-activated microfluidic cell sorting technique using an optimized, fast deep learning algorithm

Image-activated cell sorting is an essential biomedical research technique for understanding the unique characteristics of single cells. Deep learning algorithms can be used to extract hidden cell features from high-content image information to enable the discrimination of cell-to-cell differences in image-activated cell sorters. However, such systems are challenging to implement from a technical perspective due to the advanced imaging and sorting requirements and the long processing times of deep learning algorithms. Here, we introduce a user-friendly image-activated microfluidic sorting technique based on a fast deep learning model under the TensorRT framework to enable sorting decisions within 3 ms. The proposed sorter employs a significantly simplified operational procedure based on the use of a syringe connected to a piezoelectric actuator. The sorter has a 2.5 ms latency. The utility of the sorter was demonstrated through real-time sorting of fluorescent polystyrene beads and cells. The sorter achieved 98.0%, 95.1%, and 94.2% sorting purities for 15 μm and 10 μm beads, HL-60 and Jurkat cells, and HL-60 and K562 cells, respectively, with a throughput of up to 82.8 events per second (eps).

Lab on a Chip- past, present, and future

We check in with past, present, and future Editors-in-Chief, including Andreas Manz (2001–2008), Abe Lee (2017–2020), and Aaron Wheeler (2021–), about the state of the field.

Breakthrough in purification of fossil pollen for dating of sediments by a new large-particle on-chip sorter

Particle sorting is a fundamental method in various fields of medical and biological research. However, existing sorting applications are not capable for high-throughput sorting of large-size (>100 micrometers) particles. Here, we present a novel on-chip sorting method using traveling vortices generated by on-demand microjet flows, which locally exceed laminar flow condition, allowing for high-throughput sorting (5 kilohertz) with a record-wide sorting area of 520 micrometers. Using an activation system based on fluorescence detection, the method successfully sorted 160-micrometer microbeads and purified fossil pollen (maximum dimension around 170 micrometers) from lake sediments. Radiocarbon dates of sorting-derived fossil pollen concentrates proved accurate, demonstrating the method's ability to enhance building chronologies for paleoenvironmental records from sedimentary archives. The method is capable to cover urgent needs for high-throughput large-particle sorting in genomics, metabolomics, and regenerative medicine and opens up new opportunities for the use of pollen and other microfossils in geochronology, paleoecology, and paleoclimatology.

Dissecting Response to Cancer Immunotherapy by Applying Bayesian Network Analysis to Flow Cytometry Data

Cancer immunotherapy, specifically immune checkpoint blockade, has been found to be effective in the treatment of metastatic cancers. However, only a subset of patients achieve clinical responses. Elucidating pretreatment biomarkers predictive of sustained clinical response is a major research priority. Another research priority is evaluating changes in the immune system before and after treatment in responders vs. nonresponders. Our group has been studying immune networks as an accurate reflection of the global immune state. Flow cytometry (FACS, fluorescence-activated cell sorting) data characterizing immune cell panels in peripheral blood mononuclear cells (PBMC) from gastroesophageal adenocarcinoma (GEA) patients were used to analyze changes in immune networks in this setting. Here, we describe a novel computational pipeline to perform secondary analyses of FACS data using systems biology/machine learning techniques and concepts. The pipeline is centered around comparative Bayesian network analyses of immune networks and is capable of detecting strong signals that conventional methods (such as FlowJo manual gating) might miss. Future studies are planned to validate and follow up the immune biomarkers (and combinations/interactions thereof) associated with clinical responses identified with this computational pipeline.

Super-resolution optofluidic scanning microscopy

Optofluidics enables visualizing diverse anatomical and functional traits of single-cell specimens with new degrees of imaging capabilities. However, the current optofluidic microscopy systems suffer from either low resolution to reveal subcellular details or incompatibility with general microfluidic devices or operations. Here, we report optofluidic scanning microscopy (OSM) for super-resolution, live-cell imaging. The system exploits multi-focal excitation using the innate fluidic motion of the specimens, allowing for minimal instrumental complexity and full compatibility with various microfluidic configurations. The results present effective resolution doubling, optical sectioning and contrast enhancement. We anticipate the OSM system to offer a promising super-resolution optofluidic paradigm for miniaturization and different levels of integration at the chip scale.

Deep learning-enabled imaging flow cytometry for high-speed Cryptosporidium and Giardia detection

Imaging flow cytometry has become a popular technology for bioparticle image analysis because of its capability of capturing thousands of images per second. Nevertheless, the vast number of images generated by imaging flow cytometry imposes great challenges for data analysis especially when the species have similar morphologies. In this work, we report a deep learning-enabled high-throughput system for predicting Cryptosporidium and Giardia in drinking water. This system combines imaging flow cytometry and an efficient artificial neural network called MCellNet, which achieves a classification accuracy >99.6%. The system can detect Cryptosporidium and Giardia with a sensitivity of 97.37% and a specificity of 99.95%. The high-speed analysis reaches 346 frames per second, outperforming the state-of-the-art deep learning algorithm MobileNetV2 in speed (251 frames per second) with a comparable classification accuracy. The reported system empowers rapid, accurate, and high throughput bioparticle detection in clinical diagnostics, environmental monitoring and other potential biosensing applications.

Intelligent Platelet Morphometry

Technological advances in image-based platelet analysis or platelet morphometry are critical for a better understanding of the structure and function of platelets in biological research as well as for the development of better clinical strategies in medical practice. Recently, the advent of high-throughput optical imaging and deep learning has boosted platelet morphometry to the next level by providing a new set of capabilities beyond what is achievable with traditional platelet morphometry, shedding light on the unexplored domain of platelet analysis. This Opinion article introduces emerging opportunities in 'intelligent' platelet morphometry, which are expected to pave the way for a new class of diagnostics, pharmacometrics, and therapeutics.

Image-Based Live Cell Sorting

Technologies capable of cell separation based on cell images provide powerful tools enabling cell selection criteria that rely on spatially or temporally varying properties. Image-based cell sorting (IBCS) systems utilize microfluidic or microarray platforms, each having unique characteristics and applications. The advent of IBCS marks a new paradigm in which cell phenotype and behavior can be explored with high resolution and tied to cellular physiological and omics data, providing a deeper understanding of single-cell physiology and the creation of cell lines with unique properties. Cell sorting guided by high-content image information has far-reaching implications in biomedical research, clinical medicine, and pharmaceutical development.

Analysis of signal detection configurations in optical time-stretch imaging

Optical time-stretch (OTS) imaging is effective for observing ultra-fast dynamic events in real time by virtue of its capability of acquiring images with high spatial resolution at high speed. In different implementations of OTS imaging, different configurations of its signal detection, i.e. fiber-coupled and free-space detection schemes, are employed. In this research, we quantitatively analyze and compare the two detection configurations of OTS imaging in terms of sensitivity and image quality with the USAF-1951 resolution chart and diamond films, respectively, providing a valuable guidance for the system design of OTS imaging in diverse fields.

AI on a chip

Artificial intelligence (AI) has dramatically changed the landscape of science, industry, defence, and medicine in the last several years. Supported by considerably enhanced computational power and cloud storage, the field of AI has shifted from mostly theoretical studies in the discipline of computer science to diverse real-life applications such as drug design, material discovery, speech recognition, self-driving cars, advertising, finance, medical imaging, and astronomical observation, where AI-produced outcomes have been proven to be comparable or even superior to the performance of human experts. In these applications, what is essentially important for the development of AI is the data needed for machine learning. Despite its prominent importance, the very first process of the AI development, namely data collection and data preparation, is typically the most laborious task and is often a limiting factor of constructing functional AI algorithms. Lab-on-a-chip technology, in particular microfluidics, is a powerful platform for both the construction and implementation of AI in a large-scale, cost-effective, high-throughput, automated, and multiplexed manner, thereby overcoming the above bottleneck. On this platform, high-throughput imaging is a critical tool as it can generate high-content information (e.g., size, shape, structure, composition, interaction) of objects on a large scale. High-throughput imaging can also be paired with sorting and DNA/RNA sequencing to conduct a massive survey of phenotype-genotype relations whose data is too complex to analyze with traditional computational tools, but is analyzable with the power of AI. In addition to its function as a data provider, lab-on-a-chip technology can also be employed to implement the developed AI for accurate identification, characterization, classification, and prediction of objects in mixed, heterogeneous, or unknown samples. In this review article, motivated by the excellent synergy between AI and lab-on-a-chip technology, we outline fundamental elements, recent advances, future challenges, and emerging opportunities of AI with lab-on-a-chip technology or "AI on a chip" for short.

Roadmap to Building a Cell: An Evolutionary Approach

Laboratory synthesis of an elementary biological cell from isolated components may aid in understanding of the fundamental principles of life and will provide a platform for a range of bioengineering and medical applications. In essence, building a cell consists in the integration of cellular modules into system's level functionalities satisfying a definition of life. To achieve this goal, we propose in this perspective to undertake a semi-rational, system's level evolutionary approach. The strategy would require iterative cycles of genetic integration of functional modules, diversification of hereditary information, compartmentalized gene expression, selection/screening, and possibly, assistance from open-ended evolution. We explore the underlying challenges to each of these steps and discuss possible solutions toward the bottom-up construction of an artificial living cell.