Scan it, see it, measure it, analyze it. That could be a mantra for imaging flow cytometry (IFC). This technology combines the imaging of microscopy with the high speed and high photonic sensitivity of flow cytometry. Although this technology comes with decades of hardware experience, ongoing improvements in the algorithms really expand how scientists can apply IFC.
“IFC takes an image of every cell in flow,” says Brian Hall, product manager for Amnis Imaging Flow Cytometry at Luminex. “In traditional flow cytometry, only fluorescence measurements are made with no spatial information about where that fluorescence is located within the cell.”
With IFC, combining fluorescence intensity metrics from flow cytometry with microscopic imaging allows scientists to quickly gather lots of information. “IFC is primarily used to quantify cell morphology on tens of thousands of cells,” Hall explains. “This targets applications that require information about where a given signal is located on the cells and includes experiments focused on co-localization of two probes, nuclear localization of transcription factors, measuring the apoptotic index, quantifying phases of mitosis, measuring phagocytosis of particles, quantifying changes to cell morphology, and many other microscope experiments that benefit from statistics.”
Scientists tend to turn to IFC to get more information or better information. In a CYTO University webinar on IFC, Joanne Lannigan, CEO of Flow Cytometry Support Services, says that this technology “analyzes cells in suspension in a very high-throughput manner, generally without bias.” Conversely, she indicates that fluorescent microscopy data can be biased, depending on the field of view analyzed.
In some cases, getting more from IFC means higher cell numbers or more feature information from a sample. For instance, Lannigan notes that tens to hundreds of thousands of cells can be analyzed, and some platforms can track up to 20 different fluorescent colors.
Getting the statistics right, tracking more signals, and so on, all move forward with enhancements to the IFC hardware and software.
More sophisticated software
With lasers exciting markers, optics gathering images, detectors collecting the fluorescence, and a system keeping cells moving in the device, the hardware requirements for IFC cannot be denied. For example, Hall points out that improvements in charge-coupled device (CCD) cameras impact the performance of an IFC platform. This technology platform benefits from “improvements in CCD-camera sensitivity, CCD-camera image-capture times, computer processing speeds, and data storage,” Hall notes. “IFC also benefits from enhanced microscope and flow-cytometry applications, new antibody development, and new fluorochrome development.” But, scientists can’t get much out of the raw data without more work. Fortunately, the software can facilitate analysis by making complex image processing tasks simpler, faster, and easier to understand. Plus, as software improves, scientists can do more with IFC.
“IFC benefits from improvements made to computer algorithms that can identify unique morphologies through artificial intelligence, deep learning, or segmentation and feature calculation,” Hall notes. “Most of our recent enhancements to IFC have been centered on the data-analysis side by creating new machine-learning tools.” As he adds, “This has allowed us to leverage the rapid development in artificial intelligence—AI—and apply AI to cell-based assays and biological research.”
To analyze information more completely, Lannigan points out that open-source software tools offer new opportunities. As one example, she points out work by Minh Doan—now head of imaging analytics at GSK, but at the Broad Institute of MIT and Harvard at the time of the publication.1 As Doan and his colleagues wrote: “IFC is seeing a paradigm shift from low- to high-information-content analysis, driven partly by deep learning algorithms. We predict a wealth of applications with potential translation into clinical practice.”
Focusing on phytoplankton
In fact, IFC can be used in many applications. At the Helmholtz Centre for Environmental Research, for instance, Susanne Dunker—working group head of imaging flow cytometry in the department of physiological diversity—and her colleagues have used IFC to analyze phytoplankton. Such studies can be used for basic research or other applications, such as water-quality assessment. Dunker says that they have used images acquired with an ImageStreamX Mk II—a Luminex IFC platform—“for deep-learning recognition of species and their life-cycle stage.” These scientists have also used this platform, she says, to “quantify species abundance and derive phytoplankton traits.”
Image: Imaging flow cytometry captured images of phytoplankton (left), and the data can also be used to create an image-based cytogram (right). Image and data (a modified version of that in Cytometry A. 2019. 95(8):854–868) are courtesy of Susanne Dunker.
This work depends on various advances in IFC. For example, Dunker mentions high-throughput combined with image information.2 She adds, “It is possible to use different microscopic magnifications, which makes analysis really similar to microscopic investigations.”
Like other experts noted here, Dunker states that advances from IFC require a combination of technologies. “IFC is most powerful for us as it can be combined with deep learning,” she says. “With the images trained with a deep-learning algorithm, we could not only identify phytoplankton species but also their life-cycle stage.”
Slowing down time
When asked what technologies have driven the most recent advances in IFC, Bahram Jalali, director of the photonics laboratory at the University of California, Los Angeles, says, “Faster imaging technologies.” With cells moving at 1–10 meters per second, he points out, “unless the camera has a very fast shutter speed, the image will be blurred.” To accurately image cells at high-speed flow, Jalali and his colleagues developed time-stretch IFC and combined it with deep learning.3 The time stretch microscope captures blur-free images with a shutter speed of a billionth of a second then slows the images in time so they can be detected and digitized.
“A few months ago, we reported our latest results in which the computation time was reduced from seconds to less than a millisecond,” Jalali says. “This allows the AI algorithm to classify the cells in flow and direct a cell sorter to separate the cancer cells from normal cells before the cells exit the instrument.”4 He adds, “Genetic studies, such as DNA sequencing can then be done on separated cells to identify the genetic makeup of the cancer and design a personalized treatment for the patient.”
That combination of technologies drives penetrating analysis. The hardware starts the IFC process and the software determines how much can be learned from the data. With ongoing improvements in the analysis, IFC will surely deliver advances beyond today’s vision.
References
1. Doan, M., Vorobjev, I., Rees, P., et al. Diagnostic potential of imaging flow cytometry. Trends Biotechnol. 2018. 36(7): 649–652. [PMID: 29395345]
2. Dunker, S. Hidden secrets behind dots: improved phytoplankton taxonomic resolution using high-throughput imaging flow cytometry. Cytometry A. 2019. 95(8):854–868. [PMID: 31385646]
3. Mahjoubfar, A., Churkin, D.V., Barland, S. et al. Time stretch and its applications. Nature Photonics. 2017 11(6):341–351.
4. Li, Y., Mahjoubfar, A., Chen, C.L. et al. Deep cytometry: deep learning with real-time inference in cell sorting and flow cytometry. Sci. Rep. 2019. 9(1):11088. [PMID: 31366998]