AFQ.recognition.sparse_decisions

Functions

compute_sparse_decisions(bundles_being_recognized, ...)	Compute a sparse matrix of distances to ROIs for the streamlines that are
get_conflict_count(sparse_scores)	Count how many streamlines are being considered for more than one bundle
remove_conflicts(sparse_scores, bundles_being_recognized)	Returns a dictionary of {bundle_name: np.array(accepted_indices)}

Module Contents

AFQ.recognition.sparse_decisions.compute_sparse_decisions(bundles_being_recognized, n_streamlines)

Compute a sparse matrix of distances to ROIs for the streamlines that are currently being recognized. This can be used to weight decisions by distance to ROIs, without having to create a dense matrix of distances for all streamlines and all bundles.

Parameters:

bundles_being_recognizeddict

A dictionary of SlsBeingRecognized objects, keyed by bundle name.

n_streamlinesint

The total number of streamlines in the original tractogram.

Returns:

csr_matrix

A sparse matrix of shape (number of bundles being recognized, n_streamlines), where the entry (i, j) is a score:

bundles with ROIs result in weights [2.0 to 3.0] with higher scores for streamlines closer to ROIs Non-ROI bundles result in weight 1.0 Everything else is 0.0 (implicit in sparse matrices)

AFQ.recognition.sparse_decisions.get_conflict_count(sparse_scores)

Count how many streamlines are being considered for more than one bundle

AFQ.recognition.sparse_decisions.remove_conflicts(sparse_scores, bundles_being_recognized)

Returns a dictionary of {bundle_name: np.array(accepted_indices)}