AI-Based Post-processing for Artefact Mitigation in Radiography: A Systematic Review

Area of Science:

• Medical Imaging
• Artificial Intelligence in Radiology
• Radiography Artefact Mitigation

Background:

• Projectional radiography is prone to artefacts that degrade image quality and complicate interpretation.
• Artificial intelligence (AI) offers potential solutions for post-processing artefact removal in radiographic images.

Purpose of the Study:

• To systematically review AI-based post-processing methods for artefact mitigation in radiography.
• To identify challenges hindering clinical translation of these AI techniques.
• To evaluate reporting quality using the Checklist for Artificial Intelligence in Medical Imaging (CLAIM).

Main Methods:

• Systematic literature search of multiple databases (pre-09/01/2026) for AI post-processing in radiography artefact removal.
• Inclusion of original peer-reviewed articles, excluding pre-reconstruction data methods and hardware-based approaches.
• Independent screening by two reviewers, with CLAIM assessment for reporting quality.

Main Results:

• 10 studies (2020-2025) identified from 2965 records, focusing on bone shadow, device, and general artefact removal.
• Generative Adversarial Networks (GANs) were the dominant AI architecture.
• No prospective studies were found; methodological reporting was insufficient for reproducibility, with limited CLAIM adherence (mean score 17.4/42).

Conclusions:

• The current evidence base for AI-based radiography artefact mitigation is insufficient for clinical adoption.
• Future research requires multi-institutional testing, prospective designs, and transparent reporting (open data/code).
• Focus on clinically meaningful endpoints and failure mode analysis is crucial for advancing AI in medical imaging.