Introduction to AutoNeuron

AutoNeuron quickly reconstructs neurons with process-thickness measurements. It uses an innovative set of tracing algorithms to quickly explore the entire image volume in order to identify neuronal processes and somas.

AutoNeuron creates Neurolucida models of neuronal trees as branching structures, complete with branch nodes, roots and endings. Axon and dendrite diameters are recorded at each traced point. Somas are reconstructed as a 3D volume using a set of contours.

Using proprietary algorithms, AutoNeuron performs reconstructions from multiple image modalities, such as confocal, brightfield and widefield fluorescent images and stacks.

AutoNeuron defines reconstruction as a three-stage process:

• Stage 1: Using the attributes of the image background (darker or lighter than the neurons) and the size constraints that differentiate the somas from processes, AutoNeuron makes an initial estimate.
• Stage 2: AutoNeuron detects somas as 2D regions or 3D volumes larger than the thickest process with a level of contrast relative to the surrounding background region.
• Stage 3: AutoNeuron explores image regions that potentially belong to the neurons. Each exploration begins at a seed point and ideally stops at end points. Generally, an absence of seeds causes missed branches while spurious seeds lead to unwanted background traces.

AutoNeuron detects seeds by sampling the image along uniformly spaced grids.

In automatic exploration mode, AutoNeuron begins at a seed point and determines the next point to visit, following a branch until certain stop criteria are met.

The interactive reconstruction uses one user-specified seed at a time. You click a point in the process then click another point to define either the next point to visit if you want AutoNeuron to stop there, or the local direction of the process if you want AutoNeuron to trace the entire branch as far as it will go.

If your Neurolucida installation does not include the AutoNeuron module, contact MBF Bioscience Product Support for assistance.

While AutoNeuron is computationally-intensive, current desktop systems should provide acceptable performance.

However, a more powerful processor, more RAM, and a larger display will help you be more efficient with AutoNeuron, especially if you work with very large image stacks.