Related Experiment Video
Updated: May 27, 2025
05:34
Author Spotlight: Advancing Pathogen Diagnostics with Standardized LAMP
Published on: September 8, 2023
615
Using Machine Learning and Optical Microscopy Image Analysis of Immunosensors Made on Plasmonic Substrates:
Pedro R A Oiticica1,2, Monara K S C Angelim3, Juliana C Soares1
1São Carlos Institute of Physics (IFSC), University of São Paulo (USP), São Carlos, SP 13566-590, Brazil.
ACS Sensors
|February 17, 2025
Summary
A new diagnostic platform uses optical microscopy and machine learning to detect SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2) at very low concentrations. This advanced system achieves high accuracy, potentially enabling early screening of infected patients.
Area of Science:
- Biomedical Engineering
- Machine Learning Applications
- Infectious Disease Diagnostics
Background:
- Current SARS-CoV-2 detection methods have limitations in sensitivity and speed.
- Plasmonic immunosensors offer potential for high-sensitivity biosensing.
- Machine learning can enhance image analysis for biological detection.
Purpose of the Study:
- To develop and validate a diagnostic platform for sensitive SARS-CoV-2 detection.
- To integrate optical microscopy image analysis with machine learning algorithms.
- To assess the platform's performance compared to existing methods.
Main Methods:
- Utilized an optical microscopy image analysis system coupled with machine learning (ML).
- Employed support vector machine (SVM) and MobileNetV3_small convolutional neural network (CNN) models for image classification.
- Processed images from an immunosensor on a plasmonic substrate for SARS-CoV-2 detection.
Main Results:
- Achieved detection of SARS-CoV-2 at concentrations as low as 1 plaque-forming unit (PFU)/mL.
- MobileNetV3_small CNN model attained 91.6% accuracy and 96.9% F1 score for the negative class.
- Demonstrated 1000-fold lower detection limit than localized surface plasmon resonance (LSPR) sensing.
- Binary classification achieved 96.5% accuracy for SARS-CoV-2 down to 1 PFU/mL.
Conclusions:
- The developed platform shows high efficacy in detecting low SARS-CoV-2 concentrations.
- Machine learning combined with image analysis significantly enhances biosensor performance.
- This approach holds promise for straightforward screening of newly infected patients.
