CT image denoising methods for image quality improvement and radiation dose reduction

Area of Science:

• Medical Imaging
• Radiology
• Artificial Intelligence

Background:

• Growing use of computed tomography (CT) raises radiation dose concerns.
• Low-dose CT (LDCT) requires effective noise reduction techniques.
• Deep learning (DL) shows promise for advanced CT image denoising.

Purpose of the Study:

• To provide a comprehensive review of recent deep learning-based CT denoising methods.
• To analyze critical aspects beyond performance, including training, validation, and generalizability.
• To highlight challenges in developing robust DL models for CT image denoising.

Main Methods:

• Systematic literature review of CT denoising algorithms.
• Focus on deep learning (DL) approaches for low-dose CT.
• Analysis of model training, validation, testing, generalizability, vulnerability, and evaluation.

Main Results:

• Deep learning methods offer significant advancements in CT image denoising.
• Key challenges include ensuring model generalizability and vulnerability assessment.
• Standardized evaluation metrics are crucial for comparing DL denoising techniques.

Conclusions:

• Deep learning is a powerful tool for reducing radiation dose in CT imaging.
• Further research is needed to address the specific challenges of DL model development and deployment.
• This review aims to guide future research and clinical application of DL-based CT denoising.