Asymptomatic COVID-19 CT image denoising method based on wavelet transform combined with improved PSO

Area of Science:

• Medical Imaging
• Signal Processing
• Artificial Intelligence

Background:

• Gaussian noise significantly degrades asymptomatic COVID-19 computed tomography (CT) image quality, hindering image processing.
• Early COVID-19 lesions present as low-density, small ground-glass shadows easily mistaken for noise.
• Advanced COVID-19 lesions exhibit consolidation and fibrosis with gray values similar to suspected cases, posing contrast and boundary challenges.

Purpose of the Study:

• To develop an effective denoising method for asymptomatic COVID-19 CT images.
• To address the challenges of low density, noise confusion, low contrast, and fuzzy boundaries in COVID-19 CT imaging.
• To improve the accuracy of subsequent image processing and diagnostic capabilities.

Main Methods:

• A denoising method utilizing wavelet transform with a shrinkage factor is proposed.
• The threshold decreases with decomposition scale to minimize signal point misjudgment.
• An improved particle swarm optimization (PSO) algorithm optimizes wavelet threshold function parameters for adaptive changes based on lung lobe and lesion characteristics.

Main Results:

• The proposed method significantly outperforms existing algorithms in peak signal-to-noise ratio (PSNR), signal-to-noise ratio (SNR), and mean absolute error (MSE).
• For early asymptomatic COVID-19 cases, PSNR increased by approximately 5 dB, MSE was greatly reduced, and SNR improved by about 6.1 dB compared to other methods.
• The denoising effect of the proposed method is demonstrated to be superior.

Conclusions:

• The proposed wavelet transform denoising method with adaptive thresholding effectively removes noise from asymptomatic COVID-19 CT images.
• This technique enhances image clarity, particularly for early-stage ground-glass lesions and advanced consolidation/fibrosis.
• The improved image quality facilitates more accurate diagnosis and analysis of COVID-19 CT scans.