DL-assisted self-volume-calibrating colorimetric PAAHM sensors for water surveillance

Abstract

This work introduces a colorimetric sensing platform based on sodium polyacrylate hydrogel microspheres (PAAHM) integrated with a deep learning-assisted self-volume calibration strategy for the efficient and quantitative detection of NH₄⁺, PO₄³⁻, and Fe²⁺ in water. The PAAHM are uniformly loaded with colorimetric indicators through a simple immersion process, significantly simplifying sensor fabrication and enabling rapid, consistent, large-scale production. Moreover, the PAAHM exhibits dual responsiveness-both colorimetric and volumetric-allowing for the simultaneous detection of analyte concentration changes and sample volume fluctuations. To address the limitations of conventional RGB analysis, a self-volume calibration method was implemented in conjunction with a Convolutional Neural Network (CNN) to automatically extract and model both colour and morphological features from sensor images. The results demonstrate that the CNN model achieves an R² correlation coefficient of 0.999 for concentration prediction and 100% classification accuracy. This approach offers a convenient, low-cost, and highly accurate solution for on-site environmental monitoring, presenting significant potential for broad application.