A review on Deep Learning approaches for low-dose Computed Tomography restoration

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Computed Tomography (CT) provides detailed anatomical visualization.
• Low-Dose CT (LDCT) protocols reduce radiation exposure but introduce noise and artifacts.
• Conventional restoration methods struggle with LDCT image quality.

Purpose of the Study:

• To review the role and applications of Deep Learning (DL) in Low-Dose CT (LDCT) image restoration.
• To critically analyze DL-based approaches for LDCT quality enhancement.
• To identify limitations and future research directions in DL for LDCT.

Main Methods:

• Analysis of DL architectures used in LDCT restoration.
• Evaluation of performance gains and functional requirements of DL methods.
• Examination of diverse objective functions in DL-based LDCT restoration.

Main Results:

• DL approaches offer data-driven, high-performance, and fast solutions for LDCT restoration.
• Significant improvements in signal-to-noise ratio and artifact reduction are achievable with DL.
• The study highlights the current limitations and potential of DL in this field.

Conclusions:

• Deep learning is a transformative technology for Low-Dose CT image restoration.
• Further research is needed to overcome existing limitations and optimize DL applications.
• This review provides a comprehensive overview of DL in LDCT, filling a gap in existing literature.