Sharpness-Aware Low-Dose CT Denoising Using Conditional Generative Adversarial Network.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Image Processing

Background:

• Low-dose computed tomography (LDCT) reduces radiation exposure but suffers from quantum noise, potentially impairing diagnostic performance.
• Existing image denoising techniques often introduce blurring, especially at high noise levels, compromising image quality.

Purpose of the Study:

• To develop and evaluate a novel deep learning-based approach for denoising LDCT images.
• To mitigate the trade-off between noise reduction and image sharpness in LDCT reconstruction.

Main Methods:

• A deep learning model incorporating an adversarially trained network and a sharpness detection network was developed.
• The proposed method was trained and validated using both simulated and real-world LDCT datasets.

Main Results:

• The deep learning approach demonstrated minimal resolution loss in denoised LDCT images.
• Quantitative and visual assessments confirmed superior performance compared to existing state-of-the-art denoising methods.

Conclusions:

• The proposed deep learning method effectively reduces quantum noise in LDCT images while preserving essential details.
• This technique offers a promising solution for enhancing diagnostic accuracy in radiation-restricted imaging applications.