The Metric

Glossary of Terms

Ancestor - A node on the direct path to the root.

Bifurcation - A node that has two child nodes. Where a branch splits.

Branch - The segment between two nodes.

Branch order - The number of nodes between a branch and the root node.

Child - A connected node one segment farther away from the root of the tree.

Compartment - The three dimensional tapering cylinder between two points defined by the position of the points and their radii.

Confirmed match - A confirmed pairing between a gold standard node and a test node whereby the nodes are within the Euclidean distance threshold and whereby the difference between the path lengths of the nodes to their respective ancestors (that are within the Euclidean distance threshold) is within the path length error threshold. It is possible for there to be more than one node that satisfies the criteria (see 'Selected match' for more details on narrowing the matches down to one).

Continuation (metric) - A gold standard bifurcation node that does not have a matching test bifurcation node, but that does have a matching ancestor and descendant that are connected by the same path (as determined by path length). This commonly is seen when a single terminating child branch is missed in the test reconstruction.

Continuation (SWC) - In an SWC file, or in a structure with non-bifurcating and non-terminating nodes, a continuation is a node that has one parent and one child. Usually a continuation exists due to a difference in orientation or radius relative to its parent or child.

Data set - Consists of an image stack, or list of three-dimensional voxels (X, Y, Z, intensity), as well as any associated manual reconstructions and supporting material.

Degree - The number of terminal nodes in a tree (or subtree).

Descendant - A node on any path away from the root from a given node.

Digital reconstruction- A digital representation of a cell, particularly a neuron's processes. Consists of one or more binary trees starting at the root. Each tree is represented as a set of interconnected tapering cylinders, or compartments, in a standard ASCII format (.swc).

Euclidean distance threshold - The distance between a gold standard node and a test node in Euclidean space that is allowable for the nodes to match. Due to the differing potential sources of error in the XY plane versus the Z dimension, there is a separate threshold for XY and Z dimensions. This is one of several requirements for matching.

Excess node - A test reconstruction node that has no clear basis in the gold standard reconstruction. Such a node has no match, no match for its parent, and is not within the distance threshold of an unmatched gold standard node. Terminal excess nodes count as misses with weight 1, while bifurcation excess nodes count as misses with a weight determined by the number of descendant terminal excess nodes with no matches in between.

Gold standard - A manual reconstruction of a neuron in an image stack against which an automated reconstruction of the same neuron is scored.

Match - A test node that is within the Euclidean distance threshold of the target node. Multiple matches may be found and then confirmed or rejected.

Node - A point in a tree in which something changes, such as the beginning of a tree (root), the beginning of new branches (bifurcation), or the end of a branch (termination).

Parent - The immediately connected node on the path to the tree root.

Path length error threshold - A percentage that constrains matches. After a test node within the Euclidean distance threshold of the target gold standard node is found, the XY and Z error rates are determined relative to the gold standard path length (described in greater detail in the Metric Description- 2A).

Root - The first node in a tree. For an swc file, the first node in the file whose parent id is -1.

Selected match - If only one confirmed match exists for a target node, it is the selected match. If more than one confirmed match exists, then a process is undertaken to select one of the confirmed match. The process involves descendant connectivity and proximity of the confirmed matches and the target node (as described in more detail on the Metric Description page). This selected match will thereafter not be available for matching by subsequent target nodes.

Spur - A terminal branch with a path distance shorter than some defined threshold. For some data sets, spurs are removed from scoring by the metric, with each data set having its own threshold.

SWC File - A plain text file containing nodes identified by an id, type, x coordinate, y coordinate, z coordinate, radius, and parent id. Coordinate and radius units are not specified. Comments, denoted by lines beginning with a pound symbol (#), can be used to provide additional information. See the FAQ for more details.

Target node - A gold standard node that the metric is attempting to find a match for. The node of interest.

Termination - A node with no children.

Trajectory point - The point at which a gold standard path crosses the threshold boundary of the gold standard node defining the end point of the path. A separate trajectory point is calculated for the Z dimension. If the path never leaves the threshold region, the trajectory point is set as the other end point.

Test reconstruction - An automated reconstruction that the metric scores based on a gold standard manual reconstruction. Gold standard nodes are scored by looking for matches within the test reconstruction. Additional test reconstruction nodes may be considered excess nodes and count against the score.

Weight - The relative importance of a gold standard node, determined by its degree.