Sound Waves and Electric Pulses Sort Cells Without Labels

If the cell qualifies, the sound wave fires. The single-file pre-focusing matters here: every cell must take the same predictable path through the detection zone so each one gets the same sensing-activated nudge. That consistency maximizes sorting efficiency.

What did the researchers test?

The team ran the platform against three problems of increasing difficulty.

First, sorting by size. The chip separated HeLa cells from EA.hy926 cells based on differences in their effective electrical diameter, reaching up to 98% purity.

Second, sorting by stiffness. It detected and separated red blood cells whose rigidity had been chemically altered in ways that mimic disease-related changes to the cell membrane and cytoskeleton — the internal scaffolding of a cell.

Third, the hardest test: sorting HEK293T cells engineered with light-sensitive proteins. These came in two flavors — an electrically "activated" state and a state pharmacologically blocked from responding to blue light. The two look structurally identical. They differ only in how their membranes handle electrical charge. The chip told them apart at 85.4% purity.

That third result excites Ai most. "The 98% result is an important engineering benchmark, but it was achieved with differently sized cells, which are relatively easy to distinguish," he said. "Sorting cells whose light- and drug-modulated states had no obvious size or shape difference demonstrates a more distinctive capability: we can sort cells based on a subtle functional electrical signature."

What are the limits?

The purity drop from 98% to 85.4% signals where the technology stands. Functional differences between cells are smaller and more variable than size differences.

"Drug-treated cells may respond incompletely or at different rates, so their impedance features partly overlap with those of activated cells," Ai said. "Closing the gap would require careful control of event spacing and timing, as well as better signal quality and processing."

Impedance, in plain terms, is how a cell resists and reacts to an alternating electric current — its electrical fingerprint.

These are preliminary engineering results on cultured cell lines, not clinical samples. The findings show the method works in controlled lab conditions; real patient blood is a tougher, more fragile target that the team has not yet tackled.

What comes next?

Ai sees drug screening and research workflows as the nearest applications. Researchers could probe red blood cell deformability, or enrich living cells that respond differently to a treatment affecting membrane function, then study those cells further.

Clinical use — flagging disease-associated changes in blood cells — remains a longer-term goal.

The team plans two upgrades: extending the platform to primary, more fragile, clinically relevant samples, and sharpening detection of subtle functional differences with machine learning.

"Rather than relying mainly on fixed thresholds, we could train machine-learning models to recognize patterns across signal amplitude, phase, waveform shape, and frequency-dependent features," Ai said. "The challenge is to make that identification both reliable and fast enough to trigger acoustic sorting in real time."

In short: electricity reads the cell, sound moves it, and nobody has to stain anything.