Artificial intelligence for real-time motion tracking in MRI-guided radiotherapy: A systematic review

Area of Science:

• Medical Physics
• Radiotherapy Technology
• Artificial Intelligence in Medicine

Background:

• Intrafraction motion in MRI-guided radiotherapy (MRIgRT) impacts dose delivery accuracy.
• Artificial intelligence (AI) offers potential solutions for real-time motion management.

Purpose of the Study:

• To systematically review AI-driven motion tracking approaches in MRIgRT.
• To evaluate their performance and identify barriers to clinical translation.

Main Methods:

• Systematic literature search of PubMed and Web of Science (2020-2026).
• Two-stage screening, data extraction, and methodological quality assessment using the CLAIM checklist.

Main Results:

• Twenty-eight studies included, focusing on target localization (19) and motion prediction (9).
• Many studies reported sub-2 mm localization/prediction errors and high segmentation accuracy (≥0.83).
• Significant anatomical bias (liver), limited use of public datasets, scarce external validation, and insufficient source code release were noted.

Conclusions:

• AI significantly advances motion management in MRIgRT but faces clinical translation barriers.
• Barriers include anatomical bias, data limitations, validation scarcity, and metric heterogeneity.
• Future work requires multi-center datasets, anatomy-aware models, and prospective validation.