Sensor Array-Based Electronic Nose and Its Application

The development of highly selective gas sensors for specific compounds or analytespresents significant challenges, as even advanced research and commercial products struggle to achieve the desired selectivity. In our research group, efforts have focused on QCM (Quartz Crystal Microbalance)-based sensors, yet even these lack the level of gas selectivity required for precise applications. Similarly, commercially available gas sensors based on Metal-Oxide Semiconductors (MOS) are limited by low or global selectivity. Given these limitations, we propose the conceptual framework of a sensor array-based electronic nose, which combines multiple low-selectivity gas sensors with machine learning algorithms to mimic the human biological olfactory system. This system, like the human nose, cannot selectively recognize specific gases but can be trained to identify them with prior exposure. Building on this concept, we developed an electronic nose for diverse applications in medical diagnostics and food analysis. For medical applications, we conducted a trial study screening for tuberculosis (TB), achieving a sensitivity of 95% and specificity of 82%. Additionally, our electronic nose demonstrated 95% accuracy in COVID-19 detection, and after receiving clearance for market release, pre-market tests confirmed an accuracy of 88%, with both sensitivity and specificity at 88% and 84%, respectively. In neonatal sepsis detection, our system attained a 97% accuracy rate when utilizing multilayer perceptron neural networks (MLPNN). In food analysis, our electronic nose successfully differentiated the sources of meat (beef, chicken, pork) used in shredded meat products with an accuracy of 99%. Furthermore, the system achieved an impressive 95% accuracy in distinguishing between three quality grades of black tea. We have also developed a portable electronic nose for direct field detection of Ganoderma infections in oil palm plantations, offering practical, on-site applications for agricultural monitoring. The results from our applications in medical and food industries underscore the potential of electronic noses as reliable, portable, and cost-effective analytical tools. Through advances in machine learning integration and sensor array technology, our work demonstrates that electronic noses can effectively perform in various real-world scenarios where traditional, highly selective sensors fall short.