AI-enhanced CT reconstruction for texture preservation in clinical imaging

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• AI-enhanced CT reconstruction offers noise suppression and dose reduction.
• Aggressive denoising risks distorting diagnostically relevant texture, creating a 'texture preservation paradox'.
• Clinical validation of AI approaches for texture assessment is fragmented.

Purpose of the Study:

• Systematically evaluate AI-enhanced CT reconstruction techniques focusing on texture preservation.
• Summarize clinical validation evidence for dose optimization.

Main Methods:

• Systematic review following PRISMA guidelines.
• Searched PubMed, IEEE Xplore, Scopus, and Web of Science (Jan 2020 - Nov 2025).
• Included 15 clinical studies (1847 patients) evaluating AI CT reconstruction, texture assessment, and clinical validation.

Main Results:

• AI-enhanced CT consistently improved image quality (higher PSNR, SSIM, reduced noise, better lesion conspicuity) versus conventional methods.
• Studies confirmed preservation of diagnostically relevant texture, addressing limitations of conventional metrics.
• Reader validation showed improved diagnostic confidence, inter-reader consistency, and acceptable diagnostics at reduced radiation dose.

Conclusions:

• AI-enhanced CT reconstruction demonstrates clinical utility for improved image quality and dose optimization while preserving texture.
• Heterogeneity in study design and metrics necessitates cautious interpretation.
• Standardized assessment methods are crucial for future AI CT evaluations.