A biosensor that learns from you

A sensor that doesn't know you

The glucose biosensor you wear on your arm doesn't learn from you. It measures your sugar level every five minutes and sends the raw data to the app. If the trend is disturbing, you receive an alert. But the sensor doesn't know that you always go for a walk at 2:30 p.m. He doesn't know that after a meal your metabolism reacts differently than in an average patient.

This is changing. And it is changing right now, in the laboratories of the Wrocław University of Science and Technology.

Sylwia Baluta is not a science celebrity. She is a researcher who, together with a team from Wrocław, AGH, the Łukasiewicz Research Network and Vrije Universiteit Brussel, published in January 2026 in ACS Omega a text that - if read by the right people in the Polish health care system - could change the way we diagnose diseases.

Her thesis is simple: biosensors are no longer just detectors. When integrated with machine learning algorithms, they become systems that not only measure – but understand the context of measurement. They detect patterns. They predict trends. They personalize diagnostics.

Problem? Without high-quality training data, standardization, and certification, all that intelligence stays in the lab.

Six authors. Five Polish institutions. One Belgian partner. 170 references. Zero grants - which in itself is information about the state of financing of applied research in Poland.

28 billion and three barriers

The global medical biosensor market was worth $28 billion in 2024 and is growing at an annual rate of 8%. The leaders are Abbott (FreeStyle Libre), Medtronic and Dexcom. All three are investing in AI — but mainly on the cloud data analytics side, not on the sensor side itself.

This is a loophole. And this is exactly the gap that Baluta's team is targeting.

A perspective published in ACS Omega identifies three barriers that, if overcome, will open up a market worth tens of billions: lack of high-quality training data, inter-sensor variability, and lack of clinical relevance of algorithms.

The competition doesn't sleep. Startups such as BioIntelliSense and VitalConnect have collectively raised over $500 million for wearable biosensors with an analytical layer. Nature Biomedical Engineering and Science Advances publish groundbreaking work on real-time AI-corrected sweat and tear biosensors.

Regulatorily, the European Union is introducing the AI ​​Act and MDR, which together create the most stringent legal framework for AI in medicine in the world. It's a barrier. But it is also an opportunity: whoever crosses this barrier first will set the standard.

Wrocław has the answer. There's no money

Poland has three cards in this race.

First: staff. Baluta's team is not a one-time spurt. Wrocław University of Science and Technology, AGH and the Łukasiewicz Network create an ecosystem that only lacks a commercial vehicle.

Second: the internal market. 38 million patients, growing digitization of health care, cost pressure. A biosensor that will reduce the cost of monitoring a chronically ill patient by 30% will pay for itself within a year.

Third: adjustment window. The AI ​​Act comes into force by 2027. Polish companies that will now start the MDR certification process will enter the market precisely at the moment when regulations weed out competition without compliance.

But there is also a card that Poland does not have: financing. Baluta's work - 170 references, six authors - has no grant assigned. Zero. NCBR has programs, FENG has priorities, but the grant cycle lasts 18 months. At the same time, an American startup closes its seed round in three months.

What's at stake: By 2028, three to four AI biosensor platforms will be MDR certified. Poland can be a technology provider - or a user. The difference is a financing decision that has to be made now.

Sources

Main examination:Baluta S., Suresh V., Chmielowska M., Smeesters L., Cabaj J. (2026).Improving Clinical Diagnostics and Patient Care through Artificial Intelligence and Biosensor Technologies. ACS Omega. DOI: 10.1021/acsomega.5c06072

Context:BioIntelliSense, VitalConnect, Nature Biomedical Engineering (2025), Science Advances (2025)

Regulations:EU AI Act (2024/1689), EU MDR (2017/745)