Scholarlink

<p> A&#xa0;portable electrochemical aptasensor integrated with machine learning was developed&#xa0;for rapid and on-site detection of <i>Staphylococcus aureu</i>s (<i>S. aureus</i>) in food and beverage samples. The aptasensor was fabricated using screen-printed carbon electrodes (SPCEs) modified with gold nanoparticles (AuNPs) and functionalized with an Iron-regulated Surface Determinant Protein A (IsdA)-specific aptamer for the detection of <i>S. aureus</i>. Approximately 2,000 cyclic voltammetry (CV) data points were collected for six different food and beverage matrices spiked with varying concentrations of <i>S. aureus</i> (1, 10, 500, and 1000 colony-forming unit (CFU)/mL). Each CV scan was repeated 10 times, linearly averaged, and baseline corrected before model input. Noise filtering and normalization were performed to ensure consistent feature representation across training and testing datasets. Machine learning models, including Convolutional Neural Networks (CNNs) and Transformer architectures, were applied to classify the samples. The CNN model demonstrated superior performance, with a test loss of 0.0402 and a test accuracy of 99.21%. In contrast, the Transformer model achieved a test loss of 0.2014 and an accuracy of 94.21%. To enhance usability, an Android application was developed using the Network Enabled Technologies (NET) framework, enabling real-time inference of bacterial concentration directly from CV data on mobile devices (e.g. smartphones). This system demonstrates potential for a rapid, accurate, and scalable solution for real-world food safety monitoring.</p> Graphical abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Portable AI-driven electrochemical aptasensor for real-time Staphylococcus aureus detection in food and beverages

  • 期刊论文

  • 发表日期：

    2025年10月22日

查看全文
  • Shokoufeh Soleimani,
  • Tianhao Wu,
  • Jian Liu,
  • Michael K. Danquah

摘要

A portable electrochemical aptasensor integrated with machine learning was developed for rapid and on-site detection of Staphylococcus aureus (S. aureus) in food and beverage samples. The aptasensor was fabricated using screen-printed carbon electrodes (SPCEs) modified with gold nanoparticles (AuNPs) and functionalized with an Iron-regulated Surface Determinant Protein A (IsdA)-specific aptamer for the detection of S. aureus. Approximately 2,000 cyclic voltammetry (CV) data points were collected for six different food and beverage matrices spiked with varying concentrations of S. aureus (1, 10, 500, and 1000 colony-forming unit (CFU)/mL). Each CV scan was repeated 10 times, linearly averaged, and baseline corrected before model input. Noise filtering and normalization were performed to ensure consistent feature representation across training and testing datasets. Machine learning models, including Convolutional Neural Networks (CNNs) and Transformer architectures, were applied to classify the samples. The CNN model demonstrated superior performance, with a test loss of 0.0402 and a test accuracy of 99.21%. In contrast, the Transformer model achieved a test loss of 0.2014 and an accuracy of 94.21%. To enhance usability, an Android application was developed using the Network Enabled Technologies (NET) framework, enabling real-time inference of bacterial concentration directly from CV data on mobile devices (e.g. smartphones). This system demonstrates potential for a rapid, accurate, and scalable solution for real-world food safety monitoring.

Graphical abstract