Synthetic CT for gamma knife radiosurgery dose calculation: A feasibility study

Area of Science:

• Radiotherapy Physics
• Medical Imaging
• Computational Biology

Background:

• Accurate dose calculation is crucial for Gamma Knife radiosurgery (GKRS).
• Conventional CT scans involve radiation exposure and may have limitations for dose calculation.
• Synthetic CTs (sCT) generated from MRI offer a potential alternative for dose calculation.

Purpose of the Study:

• To evaluate the efficacy of MRI-based synthetic CTs (sCT) for inhomogeneity correction in GKRS treatment planning.
• To determine if sCTs can be used for dose calculation without predefined pulse sequences.

Main Methods:

• Two sets of sCTs were generated from T1post and T2 MRI images using cycleGAN.
• Image quality of sCTs was compared to original CT (oCT) using histogram comparison, RMSE, and MAE.
• Dosimetric comparisons were performed using TMR10 and convolution algorithms on oCT and sCT.

Main Results:

• sCTs demonstrated comparable HU values to oCTs, with T2-sCT showing better performance.
• TMR10 significantly underdosed targets compared to the convolution algorithm.
• No significant difference in convolution algorithm shot time was found between oCT and T2-sCT.
• Highest dosimetric differences were observed at bone/tissue interfaces, and lowest at air/tissue interfaces.

Conclusions:

• MRI-based sCT is suitable for GKRS convolution dose calculation, potentially reducing radiation dose.
• sCTs can be generated without metal artifacts in framed cases.
• Larger, diverse datasets are recommended for improved training and clinical implementation.