Purpose: To predict real-time 3D deformation field maps (DFMs) using Volumetric Cine MRI (VC-MRI) and adaptive boosting and multi-layer perceptron neural network (ADMLP-NN) for 4D target tracking.
Methods: One phase of a prior 4D-MRI is set as the prior phase, MRIprior. Principal component analysis (PCA) is used to extract 3 major respiratory deformation modes from the DFMs generated between the prior and remaining phases. VC-MRI at each time-step is considered a deformation of MRIprior, where the DFM is represented as a weighted linear combination of the PCA components. The PCA weightings are solved by minimizing the differences between on-board 2D cine MRI and its corresponding VC-MRI slice. PCA weightings solved during the first 90s are used to train an ADMLP-NN to predict PCA weightings 120 ms ahead over the next 30s. The ADMLP-NN uses several identical multi-layer perceptron neural networks (MLP-NN) with an adaptive boosting algorithm. The predicted PCA weighting are used to build predicted 3D DFM and ultimately, predicted VC-MRIs. The method was evaluated using a 4D computerized phantom (XCAT) with patient-specific RPM curves. Effects of breathing amplitude change and ADMLP-NN parameter variation were assessed. The accuracy of the PCA curve prediction in each direction was evaluated. The predicted real-time 3D tumor was evaluated against the ground-truth using Volume Percent Difference (VPD), Volume Dice Coefficient (VDC), and Center-of-Mass-Shift(COMS).
Results: The average VPD/VDC/COMS for the predicted tumor were 17.53Â±3.81%/0.92Â±0.02/1.20Â±0.65mm, respectively, across all predicted time-steps. The correlation coefficients between predicted and actual PCA curves generated through VC-MRI estimation for the 1st/2nd principal components were 0.99/0.88, 0.99/0.63, and 0.79/0.28 in the SI, AP, and lateral directions, respectively. The prediction is relatively robust against number of neurons used in ADMLP-NN.
Conclusion: Preliminary studies showed the robustness to use artificial neural networks to predict 3D DFMs to generate predicted VC-MRI for 4D target tracking.