A hybrid deep learning approach for COVID-19 detection based on genomic image processing techniques.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• The COVID-19 pandemic necessitates rapid and accurate detection methods.
• Traditional molecular and imaging techniques for COVID-19 detection have limitations.
• Automated systems are crucial for pandemic control.

Purpose of the Study:

• To propose a hybrid approach for rapid COVID-19 detection using genomic image processing (GIP).
• To overcome limitations of existing COVID-19 diagnostic methods.
• To utilize deep learning for analyzing human coronavirus (HCoV) genome sequences.

Main Methods:

• Converted HCoV genome sequences into grayscale images using frequency chaos game representation.
• Employed AlexNet (a convolutional neural network) to extract deep features from genomic images.
• Utilized ReliefF and LASSO algorithms for feature selection, followed by KNN classification.

Main Results:

• The hybrid deep learning model achieved 99.71% accuracy, 99.78% specificity, and 99.62% sensitivity.
• The best performance was obtained using features from the fc7 layer, LASSO for selection, and KNN for classification.
• The GIP approach effectively distinguished COVID-19 from other HCoV diseases.

Conclusions:

• The proposed hybrid deep learning GIP approach offers a highly accurate and efficient method for COVID-19 detection.
• This method overcomes limitations associated with traditional diagnostic techniques.
• Genomic image processing combined with deep learning shows significant potential for infectious disease diagnostics.