Artificial intelligence-coupled paper-based detection of H2O2: an experiment customized from lab-on-paper sensing to image analysis to machine learning

Abstract

Despite the vast growth of artificial intelligence (AI) in chemical research, educational programs have struggled to keep pace in familiarizing and preparing students for its implementation. In response to this gap, we have developed an interdisciplinary experiment for senior undergraduate and postgraduate students to discuss the implementations of AI in analytical chemistry with a focus on sensing. For this, an example paper-based fluorescent sensor was employed for the detection of H2O2 using image-based signal readout and machine learning-based concentration quantification. The Horseradish Peroxidase (HRP)-catalyzed yellow fluorescence emission resulting from the interaction of ortho-phenylenediamine (OPD) with H2O2 was recorded using smartphone cameras. Various features were extracted from the images and used to develop machine learning models for quantifying H2O2 concentration. Through this experiment, an introduction to sensing and paper-based sensor development, as well as the implementation of AI in sensor data analysis, was presented. During the lab experiment, students successfully developed Artificial Neural Network (ANN) models, validated them, and optimized them to improve performance. The feedback showed that students had become familiar with AI and machine learning and had increased interest in using AI in their future research.