Paper-based SERS chip with adaptive attention neural network for pathogen identification

Area of Science:

• Biotechnology
• Analytical Chemistry
• Artificial Intelligence

Background:

• Accurate and rapid pathogen identification is crucial for public health and patient care.
• Existing artificial intelligence (AI)-assisted surface-enhanced Raman scattering (SERS) biosensors face challenges with accuracy and limited bacterial fingerprint diversity.
• There is a need for improved methods for prompt and reliable pathogen discrimination.

Purpose of the Study:

• To develop a novel multi-branch adaptive attention convolutional neural network (MBAA-CNN)-assisted paper-based SERS chip for pathogen identification.
• To enhance Raman spectra diversity and improve pathogen capture using a dual-function molecule, 4-mercaptophenylboronic acid (4-MPBA).
• To achieve high accuracy and reliability in discriminating various pathogens, including antibiotic-resistant strains.

Main Methods:

• Development of a paper-based SERS chip integrated with a novel MBAA-CNN.
• Utilized 4-mercaptophenylboronic acid (4-MPBA) for bacterial capture and Raman spectra enhancement (label mode).
• Employed the K-means algorithm for pathogen identification and compared performance against a label-free mode.

Main Results:

• The 4-MPBA labeled mode demonstrated significantly higher accuracy than the label-free mode.
• The MBAA-CNN achieved 98.6% accuracy for all pathogen species and 99.5% accuracy for antibiotic-resistant and sensitive strains.
• MBAA-CNN outperformed traditional machine learning models in terms of loss value, speed, and accuracy.

Conclusions:

• The developed MBAA-CNN-assisted paper-based SERS chip offers a prompt and reliable method for pathogen discrimination.
• This approach shows potential for early, culture-free diagnosis of pathogens.
• The technology could be applied to real-time monitoring of microbial contamination in aquatic environments.