Microcolumns in the Cerebral Cortex

Home
Research
background-
methods-
-*Neuron Recog.
-Measures
-3D
aging-
Download
Papers
People
Gallery

Methods on Neuronal Recognition

Semi-automatic Neuronal Detection

This method uses the relative darkness of the cell body (dense quasi-circular shape) and its size to distinguish neu- rons from glia, blood vessels, and endothelial cells. This method consists of: (i) applying a wavelet transform fil- ter (pre-processing), (ii) marking all regions that are lo- cally darker (preliminary candidates), and (iii) choosing those markings that are surrounded by enough dark area over a neu- ronal radius (final selection). Step (i), the wavelet filter, has the effect of de-emphasizing features smaller and larger than a given input length parameter, and to enhance all structures of comparable dimension as this length parameter. This length parameter is manually adjusted so that the filter enhances neurons while discarding smaller objects such as glial cells and bigger structures such as blood vessels. The range of values used is between 8 and 16 m, which is of the same order as a typical neuronal diameter. Next, in step (ii) we find the positions of the local “darkest” points (local max- ima). These are the most likely positions for neurons, but may contain false positives. The false positives are filtered out in step (iii) by using an algorithm that examines the neigh- borhood enclosed by a circle of a radius of value chosen to be the average neuronal radius, and that calculates the aver- age darkness level inside this circle in order to compare and select only those neighborhoods that “look” like typical neu- ronal bodies. It is this last step (iii) that imposes the size and neuronal “body” (darkness) criteria for the final selection of neuronal locations.

For a step by step tutorial with illustrative images on the use of this method, please go to here.



Automated Neuron Recognition Algorithm (ANRA)

This is our latest and most exiting work on neuronal recognition. For an exhaustive presentation and discussion, please go to here.



Contact Luis Cruz with questions or comments regarding this page.