Low-Dose CT Image Denoising with Improving WGAN and Hybrid Loss Function

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Image Processing

Background:

• Computed tomography (CT) uses X-ray radiation, posing potential health risks.
• Reducing radiation dose in CT leads to noisy images and compromised diagnostic performance.

Purpose of the Study:

• To develop a novel denoising method for low-dose CT images.
• To improve image quality and diagnostic performance in low-dose CT scans.

Main Methods:

• Developed a denoising framework based on an improved generative adversarial network (GAN).
• Employed a hybrid loss function incorporating adversarial, perceptual, sharpness, and structural similarity losses.
• Utilized perceptual and structural similarity losses for textural detail preservation and sharpness loss for image clarity.

Main Results:

• The proposed method effectively removes noise and artifacts from low-dose CT images.
• Experimental results demonstrate superior performance compared to state-of-the-art methods in visual effects and quantitative measurements.
• The method successfully preserves texture details and sharpens image boundaries.

Conclusions:

• The novel GAN-based denoising method offers significant improvements for low-dose CT imaging.
• This approach enhances image quality without compromising diagnostic accuracy.
• The hybrid loss function is crucial for preserving details and clarity in denoised CT images.