"In brief, we managed to find a way to establish a dialogue with an ionic solution with zinc and copper ions. Through that dialogue we were able to ask the system complex questions about its state," explains Zoran Konkoli (to the right), Associate Professor at the Electronics and Materials Systems Laboratory (EMSL) at the Department of Microtechnology and Nanoscience – MC2, who led the project.
The notion of time is extremely important in the researchers’ approach. In a nutshell, they have constructed an intelligent sensing substrate that accumulates information over time. In such a way, unrelated events that a single-instance measurement might miss are all accounted for. The method itself is a hybrid between supervised and unsupervised learning.
"We taught the system to speak in terms of "bar-codes", an electrical response pattern, related to different ionic states. What we have achieved is a prime example of what can happen when computer science (machine learning, reservoir computing, compression algorithms), physics (multiscale modelling), the science of microfluidics, and biochemistry (intrinsic of the brain biochemistry) meet", explains Zoran Konkoli.
The key idea is to operate an environment sensitive dynamical system in the reservoir computing mode and augment it with an auxiliary input channel. The concept of the auxiliary input channel is central to the approach used. Researchers refer to it as "the drive". The drive signal aids the sensing substrate in communicating the accumulated information more clearly, e.g. in the same way a prompter helps an actor in the theater.
The HUJI team designed ion-sensitive electronics components (an ion sensitive constant phase element). The Chalmers team provided the communication protocol, form of a useful auxiliary input and most importantly the bar-codes to look for, all these being essentially a library of useful voltage input-output patterns.
"What is interesting is that these were generated through extensive theoretical simulations of the system, where the parameters of the model were calibrated against separate experimental results. Once the model has been built, we could perform thousands virtual experiments. The signal libraries were built by using rather involved genetic optimization techniques," says Zoran Konkoli.
Why bother with zinc and copper ions? The choice to focus on these ions was made under the guidance of the leader of the HUJI team, Professor Shlomo Yitzchaik (to the left). As Professor Yitzchaik explains, these ions are important biomarkers for a series of neurophysiological disorders:
"Our studies showed that by monitoring zinc(II)-to-copper(II) ionic ratio in sera sample one can have useful information on the health condition of the patient. This method proved useful in monitoring the status of the neurodegeneration state of multiple sclerosis patients versus healthy ones. These sensors combined with the barcode methodology can open new avenues for the development of point-of-care sensing devices for immunological and inflammatory disorders, autism, Alzheimer’s disease, multiple sclerosis, skin diseases, and cancer that relies on neuropeptides as a recognition layer. The great future for this technology lies in wearable sensors contacting the skin for physiology-based emotion and stress detection systems. The ability to detect the biomarkers present in sweat and process the biological information with the barcode technology may lead for future wearable sensors as affective systems that will improve our quality of life," he says.
The study is a collaboration between scientists from Chalmers and Hebrew University of Jerusalem (HUJI). From Chalmers, Vasileios Athanasiou and Zoran Konkoli contributed to the theoretical part of the research. Aldo Jesorka, Professor at the Department of Chemistry and Chemical Engineering, provided a very complex microfluidic component that the HUJI team used in their experiments. The work was supported by the EU FET Open grant RECORD-IT.
Text: Michael Nystås and Zoran Konkoli
Illustration: Vasileios Athanasiou and Zoran Konkoli
Photo of Professor Shlomo Yitzchaik: Yoav Dudkevich
Photo of Associate Professor Zoran Konkoli: Michael Nystås
Zoran Konkoli, Associate Professor, Electronics and Materials Systems Laboratory, Department of Microtechnology and Nanoscience – MC2, Chalmers University of Technology, email@example.com
Read the article in IEEE Sensors Journal >>>
On Sensing Principles Using Temporally Extended Bar Codes
MORE ABOUT THE RESEARCH >>>
The study demonstrates an indirect sensing concept on the one of the most challenging sensing problems: ion detection. In general, it is very hard to measure properties of ionic systems, especially if the ions of interest occur in very low concentrations and their number fluctuates in time. In the standard sensing setup, the flow of information is linear, from the object of interest, in this case an ionic solution, to the user who observes the system. Further, a typical measurement is a single instance event, in terms of time, though a repeated subsequent measurement might give an impression of continuity.
This time, the researchers approached the problem differently. Instead of the traditional single instance "measurement" they thought more of a "dialogue" during which one interacts with the system over a longer time period. Such a temporally extended dialogue is much more informative than a single one-instance interaction. The biggest challenge was to develop a suitable language that can be used to communicate with the system of interest.
As a guiding principle, the researchers used an indirect sensing algorithm developed within the RECORD-IT project coordinated by Zoran Konkoli, the SWEET algorithm, defined by three modules: (1) a dynamical system that interacts with the environment of interest (the sensing reservoir), (2) an auxiliary input channel that can be used to increase the intelligence of the system and query the system at the same time, and (3) a simple readout layer that is used to inspect the state of the sensing reservoir.