Low-dose COVID-19 CT Image Denoising Using CNN and its Method Noise Thresholding

Area of Science:

• Medical Imaging and Instrumentation
• Artificial Intelligence in Healthcare

Background:

• The COVID-19 pandemic necessitates advanced medical diagnostics, with CT scans playing a crucial role in assessing infection severity.
• Vulnerable groups like infants, children, and pregnant women require reduced radiation doses during CT scans, which can introduce image noise.

Purpose of the Study:

• To develop and evaluate a new denoising technique for low-dose CT images specifically for COVID-19 diagnosis.
• To mitigate the risks associated with high radiation doses in medical imaging, particularly for high-risk patient groups.

Main Methods:

• A novel denoising approach combining a convolutional neural network (CNN) with a noise-based thresholding method was employed.
• The technique focuses on reducing noise in low-dose CT images while preserving essential diagnostic information.

Main Results:

• Visual evaluation and standard performance metrics demonstrated superior outcomes compared to existing methods.
• The proposed method effectively suppressed noise in low-dose COVID-19 CT images.

Conclusions:

• The developed denoising model shows significant potential for practical applications in medical image processing, particularly for noise reduction in low-dose CT scans.
• The technique aids in preserving clinical edge information, enhancing diagnostic accuracy while minimizing radiation exposure.