Scientists have created the first artificial tongue that can fully sense and identify flavors in a liquid environment.
The results explained in the Journal PNAS on July 15 could lead to automated systems for food safety and early detection of diseases through chemical analysis, researchers say.
This technology can also be integrated into lab equipment for chemical analysis of liquid samples. Researchers also see it as a step towards “neurogenetic computing.” This is an AI system that mimics the brain’s learning process.
You might like it
The artificial tongue is made from an ultra-thin sheet of carbon, which serves as a molecular filter for graphene oxide membranes, an ionic version of taste. Instead of separating large particles, these membranes slow the movement of the ions, allowing the device to identify and remember the taste placed on the device.
In a new study, the device identified four basic flavors, sweet, sour, salty and bitter, with accuracy of 72.5% to 87.5%, and had the accuracy of 96% of drinks with multiple flavor profiles, such as coffee and Coca-Cola. The higher accuracy is due to the electrical configuration of complex drink mixtures, making the system easier to identify. This is the first time a researcher has successfully combined sensing and information processing in a single wet system.
“The discovery, Yong Yan, professor of chemistry at the Chemistry Center at the National Center for Nanoscience Technology, told Live Science in an email: “Our devices work with liquids and can sense the environment and processing information, just like the nervous system does.”
Breakthroughs in processing liquid information
Previous tasting systems processed all the information about external computer systems, but the new systems introduce a large portion of all sensing and data processing within the liquid. This primarily liquid approach is more accurate because it allows the taste to be processed in the natural ion state rather than being converted to suit the processing of the drying system.
Related: Scientists have built an “electronic tongue” powered by AI
As traditional electronic components are malfunctions in liquids, researchers had to separate detection and processing functions. This breakthrough overcomes its limitations by using graphene oxide films that can detect and carry out much of the liquid-immersed information processing.
“We lack the components that can reliably perform sensing, logic processing, and neurogenetic computing in a liquid environment,” Yan said. “Our research is trying to tackle these critical issues head on.”
The artificial tongue functions by dissolving the compound in a liquid, which breaks down into ions. The ions pass through a special layer of carbon sheets that create incredibly small channels, thousands of times thinner than human hair.
This allows ions to create unique patterns that show the flavor that the initial compound represents. The system then “learns” this pattern and becomes more accurate to identify tastes with continuous use.
The key innovation lies in the way researchers slowed the movement of ions through channels, making them 500 times slower than usual. This slowdown gave me the time to “remember” each taste I encountered, with memories that lasted for about 140 seconds, not just milliseconds, depending on the thickness of the membrane.
The researchers compared the results with a recent study by Andrew Pannone and colleagues published in the journal Nature in October 2024. In this study, data from graphene-based electronic tongues was analyzed using neural networks run on traditional solid-state computers.
The system processes information in what scientists call reservoirs, allowing the system to learn flavors. Neural networks or processing parts of the system identify patterns and pass them for final processing.
“We used a simpler machine learning system to identify different flavors: some reservoir computing and some basic neural networks,” Yan explained. “Importantly, our physical devices actually did some of the computing work,” which is different from systems that rely entirely on external computers for processing.
The system gradually builds memories, just as how the brain learns to distinguish taste. With each exposure, the system is better at distinguishing similar preferences.
“It can definitely distinguish between complex flavors such as coffee, cola, and even mixtures. It matches the performance of Panon’s sophisticated neural network,” Yong said.
Medical and practical applications
This technology allows early detection of disease through taste analysis, helps identify the effects of drugs, and can help people who have lost their sense of taste due to neurological disorders or stroke.
Artificial tongues also help improve food safety testing, quality control of beverage production, and environmental monitoring of water supply. This can be done by identifying a specific flavor in the sample.
“These innovations lay a critical foundation for applications, from medical diagnosis to autonomous machines that can ‘savour’ the environment,” Yong said.
The results are promising, but Yong acknowledged that there are still important challenges. “The system is still too bulky for practical applications,” he told Live Science. “Detection sensitivity needs to be improved and power consumption is higher than we want.”
Yong remains optimistic about the timeline of improvements, however. “We have seen transformative advances in healthcare technology, robotics and environmental surveillance within the next decade, as we decipher the challenges of improving power efficiency, integrating multiple sensors, and develop compatible neural hardware.”
Source link
