MRI-based treatment planning with pseudo CT generated through atlas registration

Area of Science:

• Medical Physics
• Radiotherapy
• Image Processing

Background:

• Accurate electron density information is crucial for radiotherapy dose calculations.
• Magnetic resonance imaging (MRI) offers superior soft-tissue contrast compared to computed tomography (CT) but lacks direct electron density information.
• Generating pseudo CTs from MRI data enables MRI-based treatment planning, potentially improving accuracy and reducing radiation exposure.

Purpose of the Study:

• To assess the feasibility and accuracy of creating pseudo CTs from MRI data using atlas-based image registration for radiotherapy planning in pediatric brain tumor patients.
• To compare different atlas registration schemes for pseudo CT synthesis, including single atlas, arithmetic mean, and pattern recognition with Gaussian process (PRGP).

Main Methods:

• Pseudo CTs were generated by deforming multiple atlas CT images to match patient-specific MRI data.
• Nonlinear registration and histogram matching were employed to align and normalize atlas and patient images.
• Four schemes were evaluated: single atlas, arithmetic mean (6 atlases), PRGP (6 atlases), and PRGP (12 atlases).
• Image similarity between pseudo CTs and real CTs was assessed, and dose distributions were recalculated and compared.

Main Results:

• Atlas-based methods outperformed a simple water-equivalent bulk CT assignment.
• Multi-atlas schemes demonstrated higher similarity to real CTs (correlation coefficient 0.787-0.819) compared to single-atlas methods.
• Dose distributions calculated using pseudo CTs closely matched original CT-based doses, with over 98% of the patient volume meeting strict evaluation criteria.
• Planning target volume (PTV) dose differences were within 2% of the prescription dose (RMS < 1%).

Conclusions:

• MRI-based treatment planning using pseudo CTs generated via multi-atlas registration is a feasible and accurate approach for pediatric brain tumors.
• The arithmetic mean method using multiple atlases provides reliable dosimetric accuracy and may be preferable to PRGP due to computational efficiency.
• This technique holds promise for enhancing radiotherapy precision and patient outcomes in pediatric oncology.