Artificial Intelligence-Based Microfluidic Platform for Detecting Contaminants in Water: A Review

Yihao Zhang, Jiaxuan Li, Zhou Yu, Xu Zhang, Xianhua LiuSensors 2024 · DOI

TL;DR

This review explores the exciting synergy between artificial intelligence (AI) and microfluidic platforms for water quality monitoring. By combining the miniaturization and efficiency of microfluidics with the analytical power of AI, researchers are developing sophisticated systems for rapid, on-site detection of a wide range of water pollutants.

What Problem Does This Paper Address?

Water pollution is a critical global issue impacting human health and ecosystems. Effectively managing this problem hinges on accurate and timely detection of contaminants. Traditional water quality testing methods, while established, often rely on bulky laboratory equipment, are time-consuming, and require specialized personnel, making them unsuitable for real-time, field-based monitoring. This creates a need for more portable, efficient, and intelligent solutions.

Methods

This paper reviews the integration of two key technologies:

  • Microfluidic Platforms: These devices manipulate tiny volumes of liquids within micro-scale channels. They offer advantages like reduced sample volume, increased sensitivity, high integration, and faster analysis. The review discusses various substrate materials (silicon, quartz, glass, polymers, and paper) and detection methods (electrochemical, optical, electroluminescence, colorimetric, mass spectrometry, and SERS) commonly employed in microfluidic systems.
  • Artificial Intelligence (AI): AI, particularly machine learning (ML) and deep learning (DL), excels at pattern recognition, classification, and prediction from complex data. The paper introduces core AI concepts like supervised, unsupervised, and reinforcement learning, and delves into specific DL architectures such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs). AI is leveraged to analyze the data generated by microfluidic sensors, enabling more accurate and automated pollutant identification.

The core innovation lies in the combination of these two technologies, where microfluidics provides the sample handling and initial sensing, and AI provides the advanced data processing and interpretation.

Datasets

The paper discusses various studies that utilize datasets for training and evaluating AI models. Specific details on the size and nature of these datasets are provided for individual applications:

  • Heavy Metal Detection: A study by Pennacchio et al. [84] compiled a dataset from over 700 images of microtiter plates for training ML models.
  • Pesticide Detection: Sahin et al. [90] used 583 Raman spectra of four insecticides for training and testing.
  • Micro- and Nanoplastic Detection: Gong et al. [105] constructed a comprehensive training dataset by combining lab-collected and publicly available data. Valentino et al. [104] used fractal features from DH-wrapped phase diagrams.
  • Microalgae Detection: Zheng et al. [116] created a dataset from 812 captured images, with 630, 770, 737, and 736 labels for four specific microalgae species. Liao et al. [115] also developed a dataset for algae image acquisition and classification.

The paper highlights the challenge of obtaining large, high-quality labeled datasets for training AI models in real-world water quality analysis.

Key Insights and Ideas

  • Synergistic Power: The combination of microfluidics and AI is not just additive; it creates a powerful new paradigm for water quality monitoring, enabling real-time, on-site analysis with unprecedented accuracy and efficiency.
  • Democratizing Expertise: AI-driven microfluidic systems can empower non-expert personnel to conduct sophisticated water quality assessments, reducing reliance on highly trained technicians.
  • Beyond Simple Detection: AI can move beyond simple detection to classification, regression (quantification), and even prediction of pollutant levels and their impacts.
  • Material Versatility: The review showcases how different microfluidic substrate materials (from rigid silicon to flexible polymers and low-cost paper) can be tailored for specific applications and detection modalities.
  • Addressing Complex Pollutants: The paper demonstrates the application of this integrated approach to a diverse range of challenging pollutants, including heavy metals, pesticides, microplastics, and microalgae.

Key Results

The paper summarizes results from various studies, highlighting the performance of AI-based microfluidic systems:

  • Heavy Metal Detection: Pennacchio et al. [84] achieved detection limits of 0.4 nM for mercury (II) and demonstrated the ability of ML models to predict concentrations without traditional readers.
  • Pesticide Detection: Sahin et al. [90] reported high classification accuracies for four insecticides, with k-NN achieving up to 0.9997 for phosmet.
  • Micro- and Nanoplastic Detection: Valentino et al. [104] achieved 98.5% accuracy in classifying MPs and diatoms using SVM. Gong et al. [105] reported 93% accuracy with CNN for microplastic identification.
  • Microalgae Detection: Liao et al. [115] achieved 94.27% accuracy with a deep neural network for algae classification. Zheng et al. [116] reported 92.8% accuracy for microalgae species detection using an MSDN based on YOLOv5.

Table 1 provides a consolidated overview of these results, showing accuracies ranging from 92% to 99.3% for various pollutants.

Limitations and Future Work

The authors acknowledge several limitations and propose future directions:

  • AI Algorithm Optimization: AI algorithms require continuous optimization to ensure accuracy and reliability when handling complex data from microfluidic chips.
  • Data Acquisition Challenges: Obtaining large volumes of high-quality, labeled training data for AI models remains a significant challenge in real-world water quality analysis.
  • Cost and Accessibility: While promising, the cost of these integrated technologies can be a barrier, especially for developing countries. Future work should focus on reducing costs and increasing accessibility.
  • Smarter Algorithms: Development of more sophisticated AI algorithms specifically tailored for microfluidic data analysis is needed.
  • Device Integration: The future lies in developing fully integrated, portable devices that combine microfluidic chips with AI processing units for true on-site detection and analysis.
  • Multi-Parameter Detection: Expanding capabilities to simultaneously detect multiple pollutants in water bodies is crucial for comprehensive environmental monitoring.

Summary updated

Abstract

Water pollution greatly impacts humans and ecosystems, so a series of policies have been enacted to control it. The first step in performing pollution control is to detect contaminants in the water. Various methods have been proposed for water quality testing, such as spectroscopy, chromatography, and electrochemical techniques. However, traditional testing methods require the utilization of laboratory equipment, which is large and not suitable for real-time testing in the field. Microfluidic devices can overcome the limitations of traditional testing instruments and have become an efficient and convenient tool for water quality analysis. At the same time, artificial intelligence is an ideal means of recognizing, classifying, and predicting data obtained from microfluidic systems. Microfluidic devices based on artificial intelligence and machine learning are being developed with great significance for the next generation of water quality monitoring systems. This review begins with a brief introduction to the algorithms involved in artificial intelligence and the materials used in the fabrication and detection techniques of microfluidic platforms. Then, the latest research development of combining the two for pollutant detection in water bodies, including heavy metals, pesticides, micro- and nanoplastics, and microalgae, is mainly introduced. Finally, the challenges encountered and the future directions of detection methods based on industrial intelligence and microfluidic chips are discussed.
BibTeX
@inproceedings{zhang2024artificial,
  author = {Yihao Zhang and Jiaxuan Li and Zhou Yu and Xu Zhang and Xianhua Liu},
  title = {{Artificial Intelligence-Based Microfluidic Platform for Detecting Contaminants in Water: A Review}},
  booktitle = {Sensors},
  year = {2024},
  doi = {10.3390/s24134350},
}
Figure