A Sensor That Reads Blood Without Drawing Any

When a surgeon needs to know a patient's potassium, the sequence is: draw blood, send it to an analyser, wait. The number that comes back is accurate about the patient as they were several minutes ago. For ions that set the rhythm of the heart and the distribution of water in the body, and which can move quickly during an operation, every reading is slightly historical. The obvious fix is to put the sensor in the bloodstream and stop sampling, and the obvious fix has been defeating engineers for a long time.

A group at The Hong Kong Polytechnic University has built a fibre thin and flexible enough to sit inside a vessel and report sodium and potassium continuously. Three parallel channels filled with liquid metal carry the signal, wrapped in a layered electrode surface that does the sensing. Implanted in a rabbit's vein it tracked both ions in real time, and its readings agreed with those of a standard clinical blood-gas analyser running on drawn samples.

Why it matters: Electrolyte levels guide decisions during and after surgery, and the current workflow cannot be anything other than retrospective. The field's own history is the reason to pay attention to the engineering detail rather than the concept, which the authors state without flattery: More broadly, implantable chemical sensors have been studied for decades, yet their in vivo reliability remains limited by protein adsorption, biofouling, foreign-body responses, and signal drift.

The study, by Wu et al. at The Hong Kong Polytechnic University, appears in Science Advances. Animal work was in rabbits; human blood was tested outside the body.

The number that counts is the one that did not move

For a sensor sitting in flowing blood, the hard problem is not the first reading but the hundredth. Proteins coat surfaces, the body reacts to a foreign object, and the electrical baseline wanders, so a device can be accurate on insertion and useless an hour later. The relevant result here is therefore a negative one: the difference between MIFS and GEM 3000 did not progressively increase over the 2 h monitoring period, indicating no obvious in vivo signal drift during venous monitoring. The gap against the reference analyser stayed put instead of widening.

Blood clotting on the device is the other immediate hazard, and the authors report that These results did not indicate progressive coagulation activation within the acute monitoring window. The careful phrasing is theirs, and it is the right phrasing: no sign of clotting starting up, within a short window.

Liquid metal is what makes the mechanical half work. A conductor inside a blood vessel has to bend with the vessel, and a solid wire stiff enough to carry signal reliably is stiff enough to be a problem. A metal that is liquid at body temperature conducts without resisting deformation.

One mechanistic detail is worth noting because it tempers the sensitivity claim. The device's response is larger than classical theory predicts, and the authors attribute that amplification to charge redistribution at the interface of the carbon-nanotube-rich solid contact, while the ion-selective membrane itself behaves near-Nernstian, meaning conventionally. The extra sensitivity comes from the supporting architecture rather than the sensing chemistry, which they characterise rather than leave unexplained.

What the study can't say yet

Two hours is the whole of the in vivo evidence. Set against the applications named for it, that is a short window, and the authors say as much: For future long-term applications such as intensive care monitoring or chronic disease management, sustained operational stability is essential. They then list precisely what has not been done: Nevertheless, long-term intravascular implantation, chronic anti-biofouling performance, and multi-day calibration stability remain to be evaluated in future studies. Given that biofouling and drift are the failures that have sunk this class of device for decades, a two-hour demonstration establishes that the sensor works, not that it keeps working.

No person was instrumented. The implantation was in rabbits, whose vessels differ in size and flow from a human's, and the human component was blood in tubes rather than blood in a patient. Rabbit veins are also not the environment an intensive-care device would face, which is likely to be an artery or a central line in someone whose physiology is unstable.

The sensor reads two ions. A demonstration pairing sodium measurement with glucose suggests the architecture can host other kinds of sensing, and the authors frame extension to pH, lactate, oxygen, drugs or blood-flow measures as future expansion rather than something shown here.

Quick questions

Could this reach patients soon? Not on this evidence. It is a rabbit study over two hours with ex vivo human blood, and the durability questions that matter clinically are explicitly left open.

Why is two hours not enough? Because the known failure mode is slow. Protein deposits and baseline drift accumulate over hours to days, so the test that matters is one long enough for them to appear.

What is the one-line takeaway? A flexible liquid-metal fibre implanted in rabbit veins tracked sodium and potassium continuously for two hours without drifting away from a clinical analyser, which is the right first result for a device class repeatedly defeated by biofouling, and well short of the multi-day stability its intended uses would require.