Non-Invasive Technique-Based Novel Corona(COVID-19) Virus Detection Using CNN

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Diseases

Background:

• The emergence of SARS-CoV-2 in late 2019 highlighted gaps in understanding its epidemiology, impacting disease control.
• Accurate and rapid diagnostic tools are crucial for effective monitoring and management of novel respiratory syndromes.

Purpose of the Study:

• To develop a high-speed, accurate, and sensitive CT scan approach for diagnosing COVID-19.
• To leverage deep learning for automated analysis of CT images in identifying coronavirus infection.

Main Methods:

• Utilized the ResNet architecture, a type of Convolution Neural Network (CNN), for image analysis.
• Trained the CNN model on CT scan images to detect characteristic patterns of COVID-19, such as peripheral lung shadows and interstitial shifts.
• Validated the model's performance on a sample dataset of CT images.

Main Results:

• The proposed method achieved a diagnostic accuracy of 95.09% and a specificity of 81.89%.
• Demonstrated a sensitivity of 100% in correctly identifying COVID-19 positive patients within the tested dataset.
• The system effectively classified COVID-19 positive cases based solely on CT image analysis.

Conclusions:

• The ResNet-based CNN approach offers a promising, highly sensitive tool for rapid COVID-19 diagnosis using CT scans.
• This method can accurately classify patients with COVID-19, aiding in timely clinical decisions and public health responses.
• Further validation without external data factors like location or population density shows the model's focused diagnostic capability.