In the laboratory of Prof. Oded Shusiov in the Faculty of Agriculture of the Hebrew University, a team of researchers led by the doctoral student Vald Shomeiko developed a computerized nose that imitates the combined action of our nose and brain and knows how to identify different smells by processing and analyzing the way in which the volatile molecules bind to the receptors using machine learning
Our nose has only about 400 receptors, yet we are able to smell millions of different odors that are carried in the air in the form of volatile molecules. The receptors we have are not specific for each smell, but we can still detect a wide variety of different smells because the volatile molecules that bind to the receptors undergo processing in the brain that knows how to translate the special pattern of binding of the molecules and identify specific smells. Moreover, our brain learns and catalogs smells that we have encountered before, so we recognize certain smells even without the visuals of the smelly object. In the laboratory of Prof. Oded Shusiov in the Faculty of Agriculture of the Hebrew University, a team of researchers led by doctoral student Vald Shomeiko developed a computerized nose that imitates the combined action of our nose and brain and knows how to identify different smells by processing and analyzing the way in which the volatile molecules bind to the receptors using machine learning. The research was done in collaboration with Prof. Yossi Platiel and Prof. Zvi Hioka from the Hebrew University and Dr. Gili Bisker from Tel Aviv University.
"The intention was to develop a universal device that could smell different smells without the need to produce a separate sensor for each new smell," explained Wald. "In millions of years of evolution, nature has created amazing tools to deal with everyday challenges very effectively, that's why we decided to learn from nature. The device we developed simulates the way our nose works." The team of researchers presented for the first time ever the use of special optical properties of carbon tubes, nanometric molecules (one million times smaller than a millimeter), for the benefit of identifying volatile substances. Carbon tubes are characterized by many special properties, among them the emission of light in the infrared range. With the help of chemical modifications of carbon tubes, they can be made to bind to specific substances and react by changing the optical signal. The optical nose designed by the researchers has a sensor array consisting of carbon tubes that simulate the receptors in the nose and the binding of volatile molecules to them causes a change in the optical signal in those sensors. Thus, the smell actually becomes a picture. The received optical signal is automatically analyzed using deep learning and thus the system learns to recognize different smells and different molecules in a similar way to our brain.
"Today, the system knows how to identify dozens of different smells and the hand is still slanted," said Weld and added, "the machine is even more effective than the human nose because it can identify smells that we cannot. For example, not only will the machine know how to differentiate between the smell of beer, vodka and wine, but it will also know what the alcohol component is inside the drink; Ethanol, methanol, propanol and more, so if you were sold a fake alcoholic drink, the machine will know it in a few seconds." Prof. Shusiov continued to explain the applicability of the development: "In addition to many and varied uses in the food sector, such as detecting fake alcohol or spoiled or poisoned food, there are many other applications for this technology. During this period, most of the content in the news is about the corona disease, and as with many viral diseases of the respiratory tract, the corona virus also causes changes in the profile of volatile molecules that are emitted to patients in the breath. Potentially, the system can be used to quickly identify corona patients in a non-invasive way by analyzing their breath. These are just a few examples - the applied potential of our technology is far beyond that." The study was recently published in the journal Biosensors and Bioelectronics.
More of the topic in Hayadan:
