Local neural network for three-dimensional dose calculations
AssigneeSiemens Healthineers International AG
InventorsZheyong Fan, Petri Hiltunen, Ari Harju
Life of this patent
Estimating the term…
Loading patent…
AssigneeSiemens Healthineers International AG
InventorsZheyong Fan, Petri Hiltunen, Ari Harju
Estimating the term…
Embodiments described herein provide for training a local neural network model that receives a first data set corresponding to parameters of a radiation therapy delivery system for delivering a radiation field to a phantom. The neural network model has been trained to determine the quantity of radiation delivered to a plurality of volume elements of the phantom based on the parameters of the first data set. The local neural network model receives local dose calculation parameters, such as total energy released per mass (TERMA) values and density values, for each voxel grid element of a high resolution voxel grid of the phantom. In an embodiment, each voxel grid element includes a central voxel and a plurality of neighboring voxels. The processor applies the neural network model to the TERMA values and the density values of the neighboring voxels to determine the quantity of radiation delivered to the central voxel.
A method comprising:
A system comprising:
Dependencies are inferred from the claim text.