Feasibility of Magnetic Resonance-based Synthetic Computed Tomography for Proton Dose Calculation in Prostate Cancer

Area of Science:

• Medical Physics
• Radiotherapy
• Medical Imaging

Background:

• Proton therapy requires accurate electron density information, typically from CT scans.
• MRI offers superior soft-tissue contrast but lacks electron density data.
• Synthetic CT (sCT) generation from MRI is an emerging technique to bridge this gap.

Purpose of the Study:

• To evaluate the feasibility of deep learning-generated sCT for proton dose calculations in prostate cancer patients.
• To compare sCT with conventional CT for image quality, proton range accuracy, and dosimetric outcomes.

Main Methods:

• sCT images were generated from T2-weighted MRI using MRI Planner for 10 prostate cancer patients.
• Image quality was assessed using Hounsfield Unit (HU) differences (MAE).
• Geometric accuracy (DSC), proton range shift (RS), and dose distributions (DVH, gamma index) were compared between sCT and CT.

Main Results:

• Mean absolute errors for sCT were within acceptable ranges for body, bone, and soft tissues.
• Good geometric agreement (DSC) and clinically acceptable proton range shifts (0.02%-0.67%) were observed.
• Dosimetric evaluations showed <1.0% relative dose differences and >90% gamma pass rates.

Conclusions:

• Deep learning-based sCT from MRI is a feasible tool for proton dose calculation in prostate cancer.
• This technology supports the potential implementation of MRI-only proton therapy workflows.