How does an electronic tongue sense: part 2

Part 1 showed two production techniques for electronic- or E-tongues. Researchers are busy investigating ways to use advanced artificial intelligence (AI) in conjunction with new sensing techniques to provide even more capabilities to E-tongues.

Sensing for food problems

To detect food fraud, spoilage, and contamination within minutes, researchers at Penn State University developed their version of an electronic tongue. Made from graphene, the ion-sensitive field-effect transistor (ISFET) sensor is only one atom thick and conducts electricity very well. The data from the sensor is processed with artificial intelligence to recognize patterns, and it has proven very effective in detecting tiny differences between similar liquids.

The non-functionalized sensors (one sensor can detect different types of chemicals, rather than requiring a specific sensor dedicated to each potential chemical) can detect various substances. Initially, the neural network was trained using 20 specific parameters related to how sample liquids interact with the sensor’s electrical properties. However, when the system was allowed to develop its own parameters from the raw sensor data, it led to some surprising results. With human-selected metrics, the system’s accuracy was 80%. This progressed to more than 95% when the AI defined its own parameters for analysis.

Figure 1. A graphene-based ISFET sensor that detects chemical ions is linked to an artificial neural network (upper right-hand corner) and trained on various datasets. (Image: the Das Lab)

From a sensor standpoint, the system does not need perfectly identical sensors to function correctly. The AI can adjust for small differences between sensors, similar to how the human brain adjusts to slight variations in taste buds. This means the E-tongue could be much cheaper to produce in large quantities since manufacturers do not need to worry about making every sensor exactly the same.

Another research group reported on their efforts to improve the capability of traditional E-tongues that suffer from a bulky size and require larger sample volumes and extra power sources. This limits their in vivo medical diagnosis and analytical chemistry applications. The researchers developed a multichannel triboelectric bioinspired E-tongue (TBIET) device. Integrated on a single glass slide chip, the device’s taste classification accuracy was improved by utilizing numerous sensory signals.

Triboelectrification is a process where two originally uncharged bodies become charged when they are brought into contact and then separated. It occurs for all highly insulating materials, but its occurrence in polymers means it is capable of storing electrical charges for a long time if it is sufficiently insulated. This solves the power requirement for the sensor.

Figure 2. An integrated multichannel triboelectric bioinspired E-tongue (TBIET) device was developed on a single glass slide chip. (Image: Nature)

The detection capability of the TBIET was validated using various test samples, including representative human body, environmental, and beverage samples, where it achieved a remarkably high classification accuracy. For example, in chemical solutions, it showed 100% identification accuracy. In environmental samples, it achieved 98.3% accuracy, and with four typical teas, it demonstrated 97.0% accuracy. Also, the classification accuracy of NaCl solutions with five different concentrations reached 96.9%. As a result, the TBIET has demonstrated a high capacity to detect and analyze droplets with ultrahigh sensitivity to their electrical properties.

A taste of the future

Taste provides an essential ability for humans to enjoy good or great products and reject bad and even dangerous liquids and foods. Similarly, electronic tongues could provide valuable differentiation and treatments for food and bacteria without the need for and/or before humans come in contact with them.