2088 - A Novel Method to Generate Synthetic MRI from CBCT for Deep Inspiration Breath Hold Abdominal Treatments

Purpose/Objective(s): Synthetic MRI would be beneficial for abdominal patients treated in Deep Inspiration Breath Hold (DIBH) on a conventional Linac.

Materials/Methods: XCAT phantom & patient data were used to develop and test a patient-specific model to generate synthetic MRI (synMRI). During patient simulation, 6 DIBH MRIs and 1 DIBH CT were acquired. One DIBH MRI was set as the reference MRI (refMRI) & principal component analysis (PCA) was performed using 5 deformation field maps (DFMs) generated by applying deformable image registration between refMRI and the other DIBH MRIs. PCA eigenvalues were then sampled 1,000 times to generate new deformations applied to a synthetic CT with the same anatomic conditions as refMRI. A convolutional neural network was trained to predict the eigenvalues corresponding to on-board conditions from a CBCT to generate a final synMRI with the new DFM. Four XCAT scenarios simulated changes from simulation to treatment. The model was evaluated using mean absolute error (MAE) and root mean square error (RMSE), and the image quality was evaluated by structure similarity index metric (SSIM) and normalized RMSE (nRMSE). The accuracy of the predicted target volume for XCAT and fiducial clips for each patient was analyzed using center of mass shift (COMS) between on-board conditions (ground truth synMRI for XCAT & CBCT for patients) and the predicted synMRI. The liver dome difference was also evaluated.

Results: For model performance, MAE was 0.11±0.02 for XCAT, and 0.15±0.01, 0.15±0.02, 0.10±0.03, 0.12±0.09 for 4 patients, respectively. RMSE were 0.11±0.07, 0.16±0.03, 0.19±0.01, 0.13±0.07, and 0.14±0.04. For image quality, SSIM values were 0.999±0.001 for XCAT, and 0.998±0.001, 0.995±0.002, 0.996±0.001, 0.998±0.001 for the patients. nRMSE were 0.10±0.06, 0.58±0.03, 0.87±0.06, 0.39±0.04, and 0.35±0.02, respectively. For XCAT, the liver dome difference & tumor COMS were <0.5 mm in all scenarios. For all patients, liver dome differences were <1 mm, and fiducial COMSs were <1.2 mm.

Conclusion: The method generates synthetic-MRI based on on board conditions in DIBH abdominal treatments performed on a conventional LINAC. Results for XCAT scenarios (S_x) and each patient (P_x). S_1/S_2 is Superior Inferior amplitude increase/decrease of 1.5mm, and S_3/S_4 is Anterior Posterior amplitude increase/decrease of 1.5mm from simulation to treatment. Volume Dice Coefficient (VDC) was not calculated for patient data. (+) P_2 had only 1 fiducial clip. (*) indicates cases in which the liver dome was not included within the CBCT.