With AI-powered automated image analysis, the Attune CytPix software can do in 10 minutes what it took expert biologists 3 years of full-time analysis to unravel.

 

What new discoveries will researchers uncover?

Imagine a million white blood cells in solution, the usual number in a flow cytometry sample. If you lined up one million white blood cells end to end, the path would span roughly 25-30 feet – the length of a two-car garage, your average garden hose, a football endzone.  

Now imagine just one cell.  

At the sub-visible, micrometer scale, it’s difficult to imagine just one of anything. But a lot in life and science comes down to individual cells. The needle in the haystack. A ‘rare event’ in flow cytometry, which in everyday life we might call a cancer cell or a disease marker, by definition, occurs at a rate of <0.01% in a large cell population. For this reason, researchers’ ability to examine millions of single cells with flow cytometry is critical.  

In flow cytometry analysis, each cell is represented by one dot on a scatterplot. The problem? Dots on plots may represent cells, but they are not cells. A lot of valuable information is left off the paper.  

What could researchers understand if they could actually see their cells flicking through the laser via real-time, identifying snapshots, like high school students filing by for their yearbook portraits? And what if we had the technology not only to photograph the cells at high speeds, but to help us categorize those images automatically and identify underlying patterns?  

Now, for the first time, we do.  

At Thermo Fisher Scientific, a team of scientists has developed a first-of-its-kind automated image analysis software for the Invitrogen™ Attune™ CytPix™ flow cytometer that could promise a paradigm shift for flow cytometry. With a trove of new visual data and the aid of automation, researchers are already unlocking biological insights that had never before been in reach.  

 

Seeing is believing

In 2023, neither the Attune CytPix nor brightfield imaging in flow cytometry is actually new. For the past year, the Attune CytPix has offered high-resolution, real-time imaging of cells and been the fastest instrument on the market to do it. The Attune CytPix uses a more modern acoustic assisted focusing method to capture up to 6,000 focused, centered images per second. It also offers 4 lasers and 14 fluorescent parameters. 

It’s an exciting combination of features, though imaging on its own is a big deal for flow cytometry. Especially when, like the Attune CytPix, it does not come at the cost of normal acquisition speeds. 

“From speaking with customers, it was clear that users would really benefit from having the ability to actually see the cells of interest,” said senior staff scientist Matthew Shallice, who has worked at Thermo Fisher for more than 20 years and followed the Attune CytPix since its inception. “It’s the concept of ‘seeing is believing,’ and really breaking the paradigm of faith-based flow cytometry where for the better part of 40 years, we’ve trusted that a dot on a plot is our event of interest. With the advent of image-enhanced flow cytometry, we can now visually confirm that a cell is in fact the cell we think it is.”   

 

 

View Attune CytPix sample data and specifications at thermofisher.com/cytpix» 

 

There are certain advantages to visualizing a cell in a flow chamber as well, as opposed to the longtime alternative of plating, staining, and examining under a microscope.  

“When we’re just plating cells on a slide under a microscope, it’s very much an artificial environment where we’re flattening the cells and looking at two-dimensional images. Cells don’t grow like that in the body,” said global product manager Nicole Madfis, an immunologist and vascular biologist who until recently worked as a stem cell tissue engineer, constructing blood vessels in the lab.  

“If we’re looking at blood cells or cells that are migrating or circulating around the body in a fluid –those environments are more like what you would see in a flow cytometer, where cells can retain a spherical shape. Now we can take images of cells in something closer to their true, physiological state.”  

Still, it didn’t take long for customers to confirm the obvious – if a picture is worth a thousand words, a thousand pictures is overwhelming. With a million cells in a single flow sample, what good is even the most stunning batch of images if the average human couldn’t possibly review and sort them all?  

“Unfortunately, while brightfield imaging is valuable in terms of determining gating strategies and doing some quick quality checks of your data, you do end up with a very large data footprint without a substantial amount of numerical data to go with it,” said the project’s biology lead and flow cytometry expert Heaven Roberts. “But we know there’s just a ton of information hidden in those images, and we certainly had customers even very early on saying, ‘why don’t we have image analysis for this?’”   

 

Unlocking new layers of data

The R&D team quickly got back to work. Soon they developed a novel software that uses machine-learning technology to automatically detect objects in an image, along with morphology differences within the cell population.  

Without straying from your tried-and-true flow cytometry workflow, without any additional time, and often even without dyes or labels, this software can tell you a lot.  

With nothing more than a few clicks on the screen, you can now gather specific information about shape, texture, size, complexity; whether you have a singlet, doublet, or triplet; population differences; whether shape is compact, elongated, or spherical; the nature of cell-to-cell interactions – the list goes on.   

“With automated image analysis, instead of saying, ‘high forward scatter and side scatter,’ we can say, ‘a 14-16 micron single cell with these specific parameters. We can really narrow in and define what that dot on the plot is,” said Roberts.  

From there, in the hands of a skilled scientist, extrapolation of that quantitative data unlocks even more possibilities.  

 

 

“I might be able to determine from my data that a lymphocyte is between X and Y diameter, and I might also be able to say that it has a circularity of over 85% and that it has a skewness intensity of less than Z and so forth,” said Roberts. “And I can start to dig into some of these subpopulations of cells and really discover where it is that they are teased apart, and then start asking how my sample preparation or my panel can separate those out.”  

The revelation has Roberts buzzy with excitement.  

“I just remember how excited I was when I realized that I could separate out subsets of granulocytes based on the automated image analysis parameters, because these subsets can be very tricky to separate with reagents alone.”  

“Before I would have said, I believe these [dots] are my granulocytes, but they could just be bits of debris. I have no way to tell. They could just be very poorly stained, or maybe I didn’t add enough antibody,” Roberts said. “Well, now there’s a ground truth in the images – you can look at the images together with quantitative values from the data and you can say, for sure, that is a granulocyte. It’s like adding a few extra lasers.” 

 

Trust, but verify

This concept of ground truth is key. Shallice sees it as core to a paradigm shift that the Attune CytPix software will enable in the field of flow cytometry.  

“When we can now clearly see that those cells that we thought were singlets included doublets, and that doublets in fact turned out to be singlets, it is breaking some of those “faith-based” scatterplot paradigms and dogmas that have been central to flow cytometry. It’s transformative. We’ve put a lot of assumptions and trust into dots on plots,” said Shallice.  

His motto in flow? Trust, but verify.  

Like most scientists, Shallice doesn’t want to trade one type of ‘faith-based’ science – dots on plots – for another – an opaque, behind-the-scenes software algorithm.  

“It was important that we also added visual tools to let users see how the image processing is actually identifying cells, with the ability to turn masks on and off and features that provide progress, feedback, and notifications,” he said.  

 

 

In this way, the R&D group hopes to empower researchers to hack away at experimental uncertainty at every level, cell by cell.  

“In repeated assays, you’re often looking for small differences. You’re screening for your candidate drug, for example, or your candidate toxin. If you have two toxin candidates that are only going to cause a difference of 2%, but your gating strategy causes a difference of 3%, then how would you ever tell them apart?” said Roberts. “By adding robustness and simplicity to these gating strategies, you’re improving your statistical power. You’re improving the ability of this technology to really differentiate treatments or research outcomes in ways that you just couldn’t do before because there was too much uncertainty around the data.”  

Roberts, in particular, has confidence in this technology because she was one of the expert biologists who helped to train the software for three years full-time. She and her peers collected and analyzed millions of cells manually across a spectrum of characteristics to build out a solid foundation of ground truth.  

Now fully trained-up with the knowledge of a team of skilled biologists, the Attune Cytometric Software can do the same three years of analysis in just ten minutes.  

 

A paradigm shift in flow cytometry

One of the other amazing outcomes of this technology is maybe also one of the least expected. As it turns out, the schooling between human and machine doesn’t just go one way.  

In building a better flow cytometer, Roberts has discovered that her own expert flow skills are rapidly transforming for the better. The software algorithm is teaching her to do things differently, even on non-Attune instruments. It’s improving her gating strategy, her sample preparation, and overall time efficiency.  

“Using the Attune CytPix flow cytometer with automated imaging has fundamentally changed the way I do and teach flow cytometry,” she said. “I used to draw my single gates much more loosely and to include all these events up at the edge [of the scatterplot]. Now when I draw my single gates, even on a different machine, I draw them very, very tightly, because I know from experience in using automated analysis that there may be many aggregates in those border regions. It’s changed the way I do flow cytometry not only with this instrument, but with all the other instruments that I work on.”  

In many ways, the automated imaging software plays off the strengths of the researchers using it without requiring a burden of additional, complex training.  

“It really came down to, how do we build this automation in a way that is seamlessly integrating back into that flow cytometry experience to provide population-level statistics?” said Shallice.   

Now, the only overwhelming thing that remains is not the volume of images or the software user experience, but the potential for new insights on the horizon – the prospect of what skilled researchers might do with this technology that we haven’t even thought of yet.  

The R&D team is looking on with eager anticipation.  

“We’re finding out all sorts of ways that our customers are interested in using this tool – everything from examining drug delivering, to phagocytosis, to nerve cells, to T-cell killing assays,” said Roberts. “For example, if you have maybe a B cell and a T cell or a T cell and a tumor cell, we could actually see whether they are touching each other – is it a coincident event or a real cell-to-cell interaction? There really is no other assay to do that at this kind of speed.”  

“We’re still discovering what we can do with this. There are so many applications. What can we reveal that was previously hidden?”  

Learn more about Attune CytPix at thermofisher.com/cytpix»  

Want to read more stories like this? Subscribe to Connect to Science, your portal for life science news. 

 

## 

For Research Use Only. Not for use in diagnostic procedures.   

© 2023 Thermo Fisher Scientific Inc. All rights reserved. All trademarks are the property of Thermo Fisher Scientific and its subsidiaries unless otherwise specified.