Clinical validation of MR-generated synthetic CT by MRCAT for brain tumor radiotherapy

MRI is an emerging modality in radiotherapy (RT). Accuracy synthetic CT is the prerequisite for implementing MR-only RT planning. This study validated the commercial algorithm of MR for calculating attenuation (MRCAT) in terms of image quality and dosimetric agreement.

Brain tumor cases with 18 treated using intensity-modulated radiotherapy (IMRT) or volumetric modulated arc therapy (VMAT), and 15 treated using stereotactic radiosurgery (SRS) were analyzed. Synthetic CTs were resampled referencing planning CT. Treatment plan calculated on planning CT was recalculated on resampled MRCAT. Image quality of selected metrics and dosimetric agreements were assessed by dose-volume-histogram and 3D gamma analysis.

For IMRT/VMAT and SRS cases, mean error were 23.42 ± 1.05 and 28.39 ± 3.17 HU; mean absolute error were 38.03 ± 1.42 and 52.36 ± 2.63 HU; root mean squared error were 89.09 ± 6.65 and 108.38 ± 12.23 HU; peak signal-to-noise ratio were 29.11 ± 0.60  and 27.65 ± 0.59 dB; and structural similarity index measures were 0.88 ± 0.00 and 0.70 ± 0.01 respectively. No significant differences were identified for DVH metrics accounting the target coverage. Most OARs did not have significant dose deviation, except left lens with 0.70% higher in D-mean after recalculation (p < 0.001). For criteria of 3 mm/3%, 2 mm/2%, and 1 mm/1%, gamma passing rates for IMRT/VMAT were 99.92%, 99.42%, and 96.47%, while SRS were 99.86%, 99.52%, and 97.57% respectively. Correlation between passing rate and image quality metrics was established in IMRT/VMAT cases, with higher similarity yield better dosimetric agreement between planning and synthetic CT.

This study has validated the MRCAT for clinical use in terms of comparable image quality and dosimetric agreement with planning CT. Further case selection and MR-compatible immobilization device would be required.