Image-Based Machine Learning Using Inkjet-Printed Chemicals: Mixing Ratio Prediction and Metal Ion Detection

Area of Science:

• Analytical Chemistry
• Materials Science
• Machine Learning

Background:

• Image-based machine learning (ML) requires extensive training data.
• Traditional methods for chemical analysis can be time-consuming and expensive.

Purpose of the Study:

• To develop a rapid, low-cost method for generating training images for ML-based mixture analysis.
• To demonstrate the versatility of inkjet printing and ML in chemical sensing.

Main Methods:

• Inkjet printing of π-conjugated organic compounds to create training images.
• Development of ML models for predicting mixing ratios.
• Adaptation of the printing surface to accommodate colorless compounds.
• Application of the method for spatial mapping of Zn²⁺ ions using colorimetric sensing.

Main Results:

• Achieved ML models with mean absolute errors below 4% within hours.
• Successfully analyzed dyes with subtle color differences.
• Extended the method to colorless compounds, including isomers and macrocyclic systems.
• Enabled detection and spatial mapping of sub-microgram levels of Zn²⁺ ions without a spectrometer.

Conclusions:

• Inkjet printing provides a simple and versatile strategy for integrating π-conjugated materials with ML.
• This approach significantly accelerates the development of ML models for colorimetric sensing and mixture analysis.
• The method offers a cost-effective and accessible alternative for various chemical sensing applications.