Presentation Information
[9a-C309-4]Fabrication of machine learning based CMOS multi-ion image sensors for measuring the three major nutrients in plants
〇Rintaro Okamoto1, Yusuke Matsushita1, Ik-Hyun Kwon1, Hideo Doi1, Kotaro Takayama1, Kazuaki Sawada1, Kazuhiro Takahashi1, Yong-Joon Choi1, Toshihiko Noda1 (1.Toyohashi Univ.)
Keywords:
Agricultural Sensors,Semiconductor Image Sensors,Ionophore
This study presents a machine learning-based CMOS multi-ion image sensor for visualizing the three major nutrients in plants. A potentiometric CMOS array sensor was fabricated with three ion-selective PVC membranes targeting potassium, ammonium, and hydrogen phosphate, which were separately coated on a single sensing area. The sensor successfully exhibited potential responses corresponding to changes in ion concentrations, demonstrating the feasibility of simultaneous multi-ion detection. However, chemical crosstalk caused by interfering ions remains a challenge in multi-ion sensing systems. To address this issue, measurement data obtained from each ion-selective membrane were analyzed using Partial Least Squares (PLS) regression. The model was trained to separate the responses of individual ions and evaluated using the coefficient of determination (R²). The results showed R² values greater than 0.9 for all target ions, indicating that machine learning can effectively compensate for crosstalk and improve ion discrimination performance. These findings suggest that the proposed sensor has strong potential for in vivo monitoring of nutrient dynamics in plants. The integration of multi-ion imaging and machine learning is expected to contribute to more efficient nutrient management and further advancement of smart agriculture technologies.