Selective Electrochemical Detection of SARS-CoV-2 Using Deep Learning.

Area of Science:

• Biomedical Engineering
• Infectious Diseases
• Computational Biology

Background:

• Accurate and rapid diagnosis of COVID-19 remains a challenge despite available antigen tests.
• Large datasets are crucial for improving diagnostic accuracy and reducing false rates.
• Machine learning (ML) and deep learning (DL) offer potential solutions for analyzing complex diagnostic data.

Purpose of the Study:

• To investigate the efficacy of ML and DL algorithms in diagnosing SARS-CoV-2 using data from an ultra-fast COVID-19 diagnostic sensor (UFC-19).
• To assess the ability of these models to differentiate SARS-CoV-2 from other coronaviruses (SARS-CoV, MERS-CoV, Human CoV) and influenza.
• To evaluate the diagnostic performance metrics (sensitivity, specificity, accuracy) of the developed models.

Main Methods:

• An electrochemical sensor (UFC-19) was utilized to collect current response data from various virus samples.
• ML and DL algorithms were applied to the UFC-19 dataset to train diagnostic models.
• Convolution neural networks (CNNs) were specifically investigated for their performance in SARS-CoV-2 detection.

Main Results:

• The convolution neural networks algorithm demonstrated high diagnostic performance for SARS-CoV-2.
• Achieved sensitivity of 96.15%, specificity of 98.17%, and accuracy of 97.20% in identifying SARS-CoV-2.
• The combined DL model and UFC-19 sensor showed potential for selective SARS-CoV-2 identification within two minutes.

Conclusions:

• ML and DL approaches, particularly CNNs, can significantly enhance the diagnostic capabilities of the UFC-19 sensor.
• The developed model offers a rapid and highly accurate method for selective SARS-CoV-2 detection.
• This approach holds promise for improving COVID-19 diagnostics and reducing misdiagnosis rates.