Filters are applied to the phantoms to create a response image h[k]. This response image has the same dimensions as the phantoms (64 × 64 × 64 voxels) and is used for benchmarking. Instead of calculating one or more features on the response image using aggregation, the entire response image is to be submitted. Therefore it is important to note the following:

• The phantom data need to be converted from an integer data type to at least 32 bit floating point precision, prior to filtering.
• Apply filters using the settings provided in the instructions. Some settings refer to real-world spacing (in mm), instead of voxel spacing. Make sure that the correct boundary condition is used for each phantom.
• The response map should be exported in a (compressed) NIfTI format, with at least 32 bit floating point precision. Make sure that the response map has the correct dimensions.
• The response map should be uploaded using the field below.

The voxel-wise distance between response maps from the different teams will be compared to identify a consensus response map.

The submitted results MUST be a CSV file with the following structure (click here to download a template file), but you may leave certain configurations empty if you have not computed them.