Related Experiment Video
Updated: May 1, 2026
Microfluidic Platform with Multiplexed Electronic Detection for Spatial Tracking of Particles
Published on: March 13, 2017
Cross-reactive molecularly imprinted electrochemical sensor arrays with adaptive deep learning for multiplexed PFAS
Xingyang Cheng1, Aodong Peng1, Yu Chen1
1College of Environmental Science and Engineering, Hunan University, Changsha, 410082, PR China; Key Laboratory of Environmental Biology and Pollution Control (Ministry of Education), Hunan University, Changsha, 410082, PR China.
This study introduces a deep learning sensor array for accurately detecting multiple per- and polyfluoroalkyl substances (PFAS) in water. The novel approach enhances ultra-trace analysis, overcoming limitations of traditional methods for environmental monitoring.
Area of Science:
- Environmental Chemistry
- Analytical Chemistry
- Sensor Technology
Background:
- Per- and polyfluoroalkyl substances (PFAS) present significant water monitoring challenges due to structural similarity and ultra-trace levels.
- Molecularly imprinted polymer (MIP)-based electrochemical sensors offer potential but struggle with cross-reactivity and complex signal analysis for mixed PFAS.
Purpose of the Study:
- To develop a robust deep learning-assisted electrochemical sensing strategy for simultaneous and quantitative analysis of multiple PFAS at ultra-trace concentrations.
- To overcome limitations of traditional MIP sensors in handling cross-reactivity and nonlinear signals in complex water matrices.
Main Methods:
- Constructed a cross-reactive MIP sensor array using three electrodes templated with specific PFAS (PFBA, PFHxA, PFOA) and a universal functional monomer.
- Utilized full differential pulse voltammetry profiles to generate high-dimensional electrochemical fingerprints.
- Developed a self-adaptive convolutional-attention-multilayer perceptron (CAM) deep learning model to decode complex signals and improve learning efficiency.
Main Results:
- The deep learning framework achieved accurate and consistent quantification of multiple PFAS across a wide concentration range (0.01-1 µM).
- The proposed method demonstrated superior performance compared to conventional machine learning architectures.
- Validation in real water matrices showed high accuracy (R² > 0.87) and low error (MAE < 0.01 µM), confirming robustness.
Conclusions:
- The developed deep learning-assisted MIP electrochemical sensing strategy provides a generalizable paradigm for simultaneous quantification of low-abundance contaminants.
- This approach enhances the capability for robust environmental monitoring of PFAS in complex aquatic environments.
