‘Electronic Tongue’ Could Taste Your Drinks for You

A sensitive electronic tongue can taste when the juice starts to go wrong

Who Simon Makin edited by Sarah Lewin Frasier

The quest for an automated way to “taste” products at the speed and scale of mass production has puzzled the food and beverage industry for decades. But in a new study, researchers used machine learning to push the limits of a promising chemical sensor, meaning a robotic tongue could soon be evaluating your milk. or merlot before you do

When ions in a liquid—say, a delicious drink—touch the conductive sheet of an ion-sensitive field-effect transistor (ISFET), the electric current that flows through it varies depending on the exact composition of the liquid and the applied voltage. This allows scientists to use ISFETs to convert chemical changes into electrical signals. The chemical composition, and thus the taste, of any beverage is affected by impurities and freshness, which can be detected by ISFETs.

“The food industry has a lot of problems knowing whether food is adulterated or has something toxic,” says Saptarshi Das, an engineer at Pennsylvania State University. The first ISFETs were demonstrated more than 50 years ago, but the sensors are not widely used commercially. The advent of graphene, an ideal conductive material, helped researchers create better ISFET sensors that detect specific chemical ions. But one big problem remained: the readings varied from sensor to sensor and with changes in conditions such as temperature or humidity.

About supporting science journalism

If you like this article, please consider supporting our award-winning journalism subscribe. By purchasing a subscription, you’re helping to ensure a future of impactful stories about the discoveries and ideas that shape our world.

In natureDas and his colleagues tackled this problem by marrying ISFETs with neural networks, training a machine learning algorithm to classify drinks using sensor readings. As a result, the system can tell whether milk has been diluted, distinguish between brands of soda or coffee blends, and identify different fruit juices, while judging freshness.

During development, the team tried training based on human-selected data points, but the scientists found that nominations were more accurate if the algorithm was fed all of the device’s measurements and chose features from its own data to base its decisions on. Human-selected features were vulnerable to device changes, while the algorithm analyzed all the data at once, finding elements that changed less. “Machine learning is able to pick up subtler differences that would be difficult for humans to define,” Das explains. The system managed more than 97 percent accuracy on practical tasks.

“The data is very compelling,” says Kiana Aran, an engineer at the University of California, San Diego, who founded a company to commercialize graphene-based biosensors. Unlike the human tongue, which detects specific molecules, this type of ISFET system only detects chemicals the changes—”Limiting it to specific and predetermined chemical profiles”, such as in brand formulations or freshness areas, he says.

Das and his colleagues will then test larger and more diverse training datasets and more complex algorithms as they expand the system’s reach. For example, “you can use this technology for health applications—monitoring blood glucose levels or sweat,” says Das. “That will be another area we want to look into.”