Post-Registration Processing API

average_am_registrations(
    t_range, param_path::Dict; reg_dir_key::String="path_dir_reg_activity_marker",
    transform_key::String="name_transform_activity_marker", transform_avg_key::String="name_transform_activity_marker_avg",
    key_param_key::String="key_transform_parameters", avg_fn::Function=median
)

Averages together registrations. All regstration parameters (including image size) must be the same except for the TransformParameters.

Arguments

• t_range: Time points to register
• param_path::Dict: Dictionary containing paths to parameter files
• reg_dir_key::String (optional, default path_dir_reg_activity_marker): Key in param_path containing registration directory
• transform_key::String (optional, default name_transform_activity_marker): Key in param_path contaning transform file names
• transform_avg_key::String (optional, default name_transform_activity_marker_avg): Key in param_path contaning averaged transform file names (to be created)
• key_param_key::String (optional, default key_transform_parameters): Key in param_path containing the TransformParameters key.
• avg_fn::Function (optional, default median): Function used to average together registrations. Default median.

average_registered_images(reg_dir, central_nrrd_path, registration_problems, res_name)

Averages all registered images, assuming they are all being registered to the same time point.

Arguments:

• reg_dir: Registration directory that contains all the moving images registered to the fixed image.
• central_nrrd_path: Path to fixed image.
• registration_problems: List of registration problems of the moving images to the fixed image.
• res_name: Name of registration file (ie: resolution) to use.

make_quality_dict(problems, evaluation_functions::Dict, selection_metric::String, resolutions; mask_dir::Union{String,Nothing}=nothing)

Computes the quality of registration using NCC, nearest-neighbors distance between centroids, and manual annotation. Returns a dictionary of registration quality values for each resolution, another dictionary of the best resolution for each problem, and a dictionary of registration resolutions that failed. Outputs a text file containing registration quality values at the best resolution. It is assumed that smaller values are better for the metrics.

Arguments

• problems: list of registration problems to compute the quality of
• evaluation_functions::Dict: dictionary of metric names to functions that evaluate elastix quality on a pair of images. The evaluation functions will be given rootpath, fixed, moving, resolution, and possibly mask_dir as input, so be sure their other parameters have been initialized correctly. It is assumed that the functions output floating-point metric values.
• selection_metric::String: which metric should be used to select the best registration out of the set of possible registrations
• resolutions: an array of resolution values to be using. Each value is represented as a tuple (i,j), where i is the number of parameter file to use and j is the resolution for registrations using that parameter file. Both are 0-indexed.

Optional Keyword Arguments

• mask_dir::Union{String,Nothing}: directory to a mask file. Statistics will not be computed on regions outside the mask. If left blank, no mask will be used or passed to the evaluation functions.

make_quality_dict(param_path::Dict, param::Dict, problems, evaluation_functions)

Computes the quality of registration using NCC, nearest-neighbors distance between centroids, and manual annotation. Returns a dictionary of registration quality values for each resolution, another dictionary of the best resolution for each problem, and a dictionary of registration resolutions that failed. Outputs a text file containing registration quality values at the best resolution. It is assumed that smaller values are better for the metrics.

Arguments

• param_path::Dict: Dictionary containing path_dir_mask entry to the path of masks (or nothing if no masks are used)
• param::Dict: Dictionary containing the following keys:
  • quality_metric::String: which metric should be used to select the best registration out of the set of possible registrations
  • good_registration_resolutions: an array of resolution values to be using. Each value is represented as a tuple (i,j), where i is the number of parameter file to use and j is the resolution for registrations using that parameter file. Both are 0-indexed.
• problems: list of registration problems to compute the quality of
• evaluation_functions::Dict: dictionary of metric names to functions that evaluate elastix quality on a pair of images. The evaluation functions will be given rootpath, fixed, moving, resolution, and possibly mask_dir as input, so be sure their other parameters have been initialized correctly. It is assumed that the functions output floating-point metric values.

calculate_ncc(moving, fixed)

Computes the NCC of two image arrays moving and fixed corresponding to a registration.

metric_tfm(ncc; threshold=0.9)

Applies a function to ncc to make it a cost that increases to infinity if ncc decreases below threshold (default 0.9)