Brain-inspired technology is revolutionising cell sorting: The story of NEVACS

Imagine standing by a river and trying to identify every single leaf that floats past you—not just its colour, but its shape, size, and whether it has tiny spots or tears. Now, imagine doing this a thousand times every second. That’s essentially what scientists do when sorting cells, except instead of leaves, they’re examining microscopic biological units, each carrying critical clues about health, disease, or potential new treatments.

For decades, scientists have relied on traditional cell-sorting machines, which are like high-speed cameras, snapping a quick photo of each passing cell. But here’s the problem: a single photo can’t tell the whole story. It’s like identifying a bird species from one blurry image rather than watching a video of it flying, chirping, and behaving naturally.

This is where NEVACS—Neuromorphic-Enabled Video-Activated Cell Sorting—enters the picture. NEVACS is not just a new cell-sorting machine; it’s a leap forward, combining event cameras, brain-inspired computing, and microfluidic technology to create a smarter, faster, and more insightful way to understand cells.

NEVACS integrates event cameras, Spiking Neural Networks, and neuromorphic chips to enable educing-redundancy-for-efficiency strategy that surpasses the limitations of traditional image-activated cell sorting.

Wenhui Wang

How NEVACS sees cells in motion: The magic of event cameras

At the heart of NEVACS is something called an event camera—a technology that might sound futuristic but is already making waves in areas like robotics and self-driving cars. Let’s compare it to a regular camera to understand how it works.

A traditional camera captures a full image (or “frame”) every time the shutter clicks. Whether something in the image moves or stays completely still, the camera records every pixel equally. Now, imagine using this method to monitor traffic on a highway. Every car, every tree, and every speck of the road would be recorded in every frame, creating an enormous amount of data—even if most of those pixels didn’t change. An event camera, however, works completely differently.

Instead of capturing entire frames at regular intervals, it only records changes in brightness at each pixel. Think of it like a stadium full of fans holding flashlights. A regular camera would capture every flashlight, whether on or off, in every single frame. But an event camera would only record when a flashlight turned on, turned off or moved.

For cell sorting, this approach is a game-changer. As cells flow through the tiny microfluidic channels in NEVACS, the event camera doesn’t waste time capturing every still pixel of the background or the unmoving fluid. Instead, it focuses only on what matters: the edges of the cells, their movements, and their changing shapes as they roll and tumble through the channel.

This ability to focus on “events” instead of frames allows NEVACS to operate at blazing speeds and produce far less redundant data. It’s like listening to someone tell you only the interesting parts of a long story instead of every single word they say. And because the data is already simplified, it can be processed much more quickly—a crucial advantage when you’re trying to sort 1,000 cells per second. But video alone isn’t enough. Even with this precise, efficient visual data, NEVACS needs to decide what it’s looking at—and do so instantly.

The brain behind NEVACS: Spiking neural networks

Once NEVACS captures a cell’s video, it needs to decide what it’s seeing—and it needs to do so almost instantly. Enter the Spiking Neural Network (SNN). Think of this as a tiny artificial brain designed to work like ours. Similar to the difference between an event camera and a traditional camera, SNN can process non-static inputs compared to traditional artificial neural networks, and only important information is transmitted between its neurons in an event-driven and spiking manner. This pattern is naturally suited to processing event data and further keeps the advantages.

In NEVACS, the SNN analyses the video footage of each cell, looking for patterns, shapes, and movements that indicate whether a cell is healthy, diseased, or something else entirely. This neural network runs on a specialized computer chip called a neuromorphic chip—essentially a tiny processor optimized for brain-like thinking. Traditional computer systems process multiple parallel tasks synchronously, like a choir where everyone must be in step; otherwise, it will be messy.

But neuromorphic chips with SNN are more like a busy kitchen where everyone—chefs, waiters, and dishwashers—is working together at the same time, each handling their own task but coordinating seamlessly. The result? NEVACS can analyse and classify up to 1,000 cells per second, all while consuming far less power than a traditional computer system.

A real-world example: Detecting diseased blood cells

Let’s see NEVACS in action. Imagine you’re a doctor trying to identify diseased red blood cells (RBCs) in a patient’s blood sample. Healthy RBCs are shaped like smooth, flexible discs, while diseased ones (like those in hereditary spherocytosis) are more spherical. From certain angles, these two types of cells might look almost identical in a single snapshot. That’s why traditional sorting systems often misidentify them.

But NEVACS doesn’t rely on a single snapshot. It captures a video of each cell as it rolls and tumbles through the channel. This rolling motion gives NEVACS multiple perspectives—like examining a sculpture from every angle instead of just one side. Using this richer data, NEVACS can accurately tell a healthy RBC from a diseased one, achieving an error rate of just 0.99%, compared to nearly 20% in traditional systems. In medical diagnostics, that level of accuracy can mean the difference between catching a disease early and missing it entirely.

More than just blood: NEVACS’ versatility

NEVACS isn’t limited to blood cells—it’s like a microscope on steroids, capable of sorting everything from living and dead cells to tiny plastic particles in water samples. For example, scientists often use fluorescent dyes to mark dead cells, which glow brightly under certain types of light. NEVACS can capture both bright-field and fluorescent signals simultaneously. This means it can spot if a cell is alive or dead and analyse its shape, size, and structure—all in one go.

What’s more impressive is that NEVACS does all this using a single event camera and a simple microfluidic channel. Older systems often require multiple cameras, complex hardware, and intensive computing setups. NEVACS strips away that complexity, making it both cost-effective and easy to scale up.

Why NEVACS matters for the future

The implications of NEVACS stretch far beyond the lab. Medical research could help doctors identify diseases earlier and more accurately. In drug development, it could monitor how experimental treatments affect individual cells in real time. In environmental science, it could filter harmful microplastics from water samples with unmatched precision.

And NEVACS isn’t stopping here. Researchers are already working on improvements: more powerful neuromorphic chips, higher-resolution event cameras, and even larger systems capable of analysing multiple channels at once.

A new perspective on tiny worlds

At its core, NEVACS represents a fundamental shift in how scientists approach cell sorting. It’s not just about freezing a moment in time with a snapshot—it’s about watching cells in action, understanding their behaviour, and making smart decisions based on dynamic information.

By combining the sensitivity of event cameras, the intelligence of brain-inspired neural networks, and the efficiency of microfluidic systems, NEVACS is setting a new standard. In a world where microscopic details can carry life-saving information, technologies like NEVACS remind us that sometimes, the answers aren’t found in a single frame but in the story that unfolds over time. With NEVACS, we’re not just looking at cells anymore—we’re learning from them, one tiny movement at a time.

🔬🧫🧪🔍🤓👩‍🔬🦠🔭📚

Journal reference

He, W., Zhu, J., Feng, Y., Liang, F., You, K., Chai, H., … & Wang, W. (2024). Neuromorphic-enabled video-activated cell sorting. Nature Communications, 15(1), 10792. https://doi.org/10.1038/s41467-024-55094-0