FCOD: Fast COVID-19 Detector based on deep learning techniques.

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Deep Learning

Background:

• The COVID-19 pandemic presents a significant global health crisis, necessitating rapid diagnostic tools.
• Limitations in existing test kits highlight the need for efficient screening methods.
• Early detection of COVID-19 is crucial for preventing community transmission.

Purpose of the Study:

• To introduce a novel, automated detection system for COVID-19.
• To develop a fast and accurate diagnostic model using chest X-ray images.
• To reduce the burden on healthcare systems through rapid patient screening.

Main Methods:

• A deep learning model, Fast COVID-19 Detector (FCOD), was developed.
• FCOD utilizes an Inception architecture with 17 depthwise separable convolution layers.
• The model was trained and evaluated on the covid-chestxray-dataset comprising 940 chest X-ray images.

Main Results:

• The FCOD model achieved high performance metrics: 96% accuracy, 96% F1-score, and 0.95 AUC.
• Detection was performed rapidly, with each case classified in approximately 0.014 seconds.
• Depthwise separable convolutions contributed to reduced computational costs and model parameters.

Conclusions:

• The Fast COVID-19 Detector (FCOD) demonstrates significant potential for rapid COVID-19 diagnosis.
• The model can serve as an effective decision-support tool for radiologists.
• Immediate patient screening using FCOD can aid in controlling the spread of COVID-19.