Deep Convolution Neural Network (DCNN) Multiplane Approach to Synthetic CT Generation From MR images—Application in Brain Proton Therapy

The paper describes a multiplane deep convolution neural network approach to predict synthetic computed tomography from T1-weighted magnetic resonance imaging. The method was tested in the framework of brain proton therapy, where Hounsfield Unit inaccuracies and steep density gradients easily lead to range shift errors. Results proved that the predicted synthetic computed tomography can be suitable for proton beam planning in the vision of magnetic resonance imaging-only proton therapy.