Quality evaluation of synthetic CT using a gamma analysis for prostate cancer MR-only radiotherapy

This study evaluates the quality of synthetic computed tomography (sCT) images for MR-only radiotherapy in prostate cancer using gamma analysis. A software tool, MRGamma, was developed to address challenges like the absence of electron density maps and registration uncertainties between magnetic resonance imaging (MRI) and planning CT (pCT).

Aplication developed in MATLAB assesses Hounsfield units (HU) discrepancies between CT and sCT images via 2D and 3D gamma analysis (GA). GA computes a gamma index (γ) for each voxel based on HU differences and distance to agreement (DTA). The study analyzed prostate cancer patients using the MRCAT algorithm to generate sCT images. Preprocessing included registration, resampling, and cropping to standardize image dimensions. GA was performed with thresholds of 100 HU and 70 HU for HU differences and 3 mm for DTA to evaluate image quality.

Aplication demonstrated high agreement between sCT and CT images. For 3D GA (100 HU, 3 mm), 95.5 ± 1.8% of voxels passed the threshold, with slightly lower rates for 70 HU. Discrepancies were most pronounced at tissue interfaces and air cavities, where HU variations were more significant. Mean γ values for 3D GA were 0.19 ± 0.04 (milder parameters) and 0.23 ± 0.05 (stricter parameters), showing consistent trends in 2D GA. Maximum γ values confirmed good overall agreement.

Aplication effectively evaluates sCT quality, supporting the feasibility of MR-only radiotherapy. By providing detailed HU comparisons, the tool enhances MR-based treatment planning, reducing costs and ionizing radiation exposure. Its implementation may improve workflow efficiency and patient safety in clinical practice.