There are two ways to implement automatic tracing:
To continue incomplete automatic tracings, you can use the user-guided tracing mode.
To process multiple files with the configuration used for a representative image stack, use the batch trace functionality.
When working with multiple images, make sure that the image you want to use for tracing is checked in the Image Organizer.
With this method, four directional kernels are matched to the image data. For a given point within the dendrite/axon, the algorithm identifies the best positions and orientations for the top, bottom, left, and right kernels surrounding the point. The positions and orientations results are combined to estimate the next point to trace. Points are estimated until a set of stopping criteria is met.
The same algorithm was used in AutoNeuron. For details on the algorithm, see Rapid automated three-dimensional tracing of neurons from confocal image stacks (Al-Kofahi, Lasek, Szarowski, Pace, Nagy, Turner, and Roysam, 2002).
The detection is implemented from a branch centerline to its outer edges (manuscript in preparation).
This algorithm generates clusters of voxels iteratively along the dendrite/axon. These clusters are then used to position the nodes that define the centerline of the dendrite/axon.
The same algorithm is used in NeuronStudio (developed by the Mount Sinai School of Medicine). For details on the algorithm, see Three-Dimensional Neuron Tracing by Voxel Scooping (Rodriguez, Ehlenberger, Hof, & Wearne, 2009).