Image-based retrospective 4D MRI in external beam radiotherapy: A comparative study with a digital phantom

Area of Science:

• Medical imaging
• Radiotherapy physics
• Image processing

Background:

• Respiratory motion significantly impacts radiotherapy accuracy.
• Four-dimensional magnetic resonance imaging (4D MRI) is crucial for motion reconstruction.
• Optimal retrospective sorting methods for artifact-free 4D MRI are not yet established.

Purpose of the Study:

• To proactively compare three image-based retrospective sorting methods for 4D MRI.
• To establish a groundwork for benchmarking 4D MRI sorting strategies.
• To identify the most accurate method for respiratory motion reconstruction in radiotherapy.

Main Methods:

• Compared three published image-based sorting methods (body area, mutual information, navigator slice) against a conventional external surrogate method.
• Utilized an MRI digital phantom derived from XCAT CT with patient-derived signals (12 cases).
• Simulated multislice MRI acquisitions across sagittal, coronal, and axial orientations, generating an average 4D cycle as ground truth.

Main Results:

• Navigator slice and mutual information methods demonstrated superior amplitude binning performance over the body area method.
• The navigator slice method reduced diaphragm artifacts by utilizing internal anatomical information.
• Mutual information with amplitude binning accurately matched tumor motion ground truth; body area method showed larger errors.
• Sagittal and coronal orientations yielded lower errors than axial slicing.
• Patient-specific evaluation is necessary for optimal method selection.

Conclusions:

• The study provides a comparative analysis of image-based retrospective sorting methods for 4D MRI.
• Findings offer guidelines for defining benchmarks in MRI-guided radiotherapy.
• Navigator slice and mutual information methods are promising for accurate respiratory motion reconstruction.