AUTOMATED CORRECTION OF METAL AFFECTED VOXEL REPRESENTATIONS OF X-RAY DATA USING DEEP LEARNING TECHNIQUES

摘要

A computer-implemented method for correction of metal affected voxel representations of x-ray data is described wherein the method comprises: a first 3D deep neural network receiving a first voxel representation of metal affected x-ray data at its input and generating voxel identification information at its output, the voxel identification information identifying voxels of the first voxel representation that belong to a region of voxels that are affected by metal; a second 3D deep neural network receiving the first voxel representation and the identification information generated by the first 3D deep neural network at its input and generating for each voxel of the first voxel representation identified by the voxel identification information a predicted voxel value at its output, the 3D deep neural network predicting the predicted voxel value on the basis of training data that include voxel representations of clinical x-ray data; and, determining a corrected first voxel representation by replacing voxel values of voxels of the first voxel representation that are identified by the voxel identification information as belonging to a region of voxels that are affected by the metal.