[MRI-based radiotherapy planning].

Area of Science:

• Medical Imaging
• Radiation Oncology
• Radiotherapy Physics

Background:

• The integration of MRI with linear accelerators for radiotherapy necessitates accurate tissue density information for dose calculations.
• Current MRI technology lacks inherent electron density data required for precise radiotherapy planning.

Purpose of the Study:

• To review and compare different strategies for generating synthetic CT (pseudo-CT) images from MRI data for radiotherapy planning.
• To evaluate the dosimetric accuracy, speed, automation, and clinical applicability of various pseudo-CT generation methods.

Main Methods:

• Literature review of strategies including bulk density mapping, machine learning-based approaches, atlas-based methods, and physics-based approaches.
• Analysis of methods generating synthetic CT (pseudo-CT) and those calculating dose directly from MRI.
• Comparison based on dosimetric accuracy, computational speed, and level of automation.

Main Results:

• Atlas-based and machine learning-based approaches demonstrate superior dosimetric accuracy compared to bulk density mapping.
• Bulk density methods are unsuitable for accurate bone density representation.
• Machine learning methods are fastest, while atlas-based methods are slowest; physics-based approaches are promising.

Conclusions:

• Atlas- and machine learning-based methods are currently the most accurate for pseudo-CT generation in MRI-based radiotherapy planning.
• Physics-based approaches represent a promising alternative for direct dose calculation from MRI data.
• Rigorous validation of all methods is essential for clinical implementation, especially for adaptive radiotherapy.