The sensor that thinks for itself
EU-funded researchers have demonstrated a new type of biomedical sensor featuring embedded intelligence. The sensor measures ion concentrations but the computing principles have far-reaching implications for all kinds of biomedical and environmental sensing.
© maxsim #175806432, source: stock.adobe.com 2019
As biomedical sensors become commonplace and provide growing volumes of data, the problem of extracting useful information in real time becomes ever-more pressing. One approach is to endow the sensor with its own processing power, a concept known as embedded computing. So, how can compact but intelligent biomedical sensors be designed?
To solve such a problem, you need an intelligent sensing substrate that collects and processes information and operates on low power, says Zoran Konkoli of Chalmers University of Technology in Gothenburg, Sweden. The RECORD-IT project addressed this broad challenge of building such intelligent sensing substrates.
As coordinator of the EU-funded project, Konkoli developed a new algorithm called SWEET which he describes as an algorithmic template, a language, or a user manual for building complex sensing systems.
It is a way of designing intelligent sensing substrates that can be used for assessing the state of any complex environment with which it is hard to interact and for which the dynamics are not known, he says. An ionic solution is a prime example of such a system.
In biological fluids, many different ions may be present in varying proportions but in low concentrations and there is no simple method for measuring the ionic composition and tracking its changes. Partners in the three-year project linked several types of sensor to construct a prototype network which was able to take care of the processing itself.
We tried to combine organic and inorganic devices that interact with ions, Konkoli explains. Imagine a neural network in which each neuron is replaced by such a device. The idea is that the intelligence of the whole system should be smarter than the sum of the intelligences of the individual components. The computing power comes from interactions between the components.
RECORD-IT finished in August 2018 but the partners work on intelligent biomedical sensors is continuing. A Swiss start-up company, MOMM Diagnostics, is applying the insights from the project to the early diagnosis of pre-eclampsia in an EU-funded project called PEDPOC. Chalmers University and the Hebrew University of Jerusalem are developing a method for diagnosing neurological conditions by sensing the balance of zinc and copper ions in body fluids.
However, ion sensing is only a demonstration of an underlying computing principle that Konkoli believes will have wide-ranging and unpredictable implications, especially as information exchange intensifies between objects as well as people in the Internet of Things.
The RECORD-IT sensors embody a concept known as reservoir computing any complex dynamical system that interacts with its environment will remember traces of those interactions within it. So the computer and the sensor are one and the same.
He admits the idea is complex and not easy to explain. People do not always understand it but the potential is enormous. We have tried several applications on various problems that are miles away from ionic sensing.
For example, the SWEET algorithm could be used to design a novel approach to encryption, to help patients recover from spinal cord injuries, or even to infer thoughts from brain-activity sensors.
On a larger scale, Konkoli points out that the mobile phone network constitutes a reservoir computer that, through its many users, is sensing valuable information about the state of the world. The only thing we need to do is to query that network to figure out about the environment in which people are using their phones.