The data is obtained from a confocal microscope at about 240x
magnification.
Last update: Dec. 5, 2003
Dataset graciously shared by Prof. Mark F. Bear, Prof. Anna Dunaevsky and Wes Wallace from the Department of Neuroscience @ Brown University.
Dendritic spines are minute protrusions which cover the dendrites of the pyramidal neurons of cerebral cortex. The pyramidal neurons are themselves the principal integrative elements in cortical networks. Each dendritic spine is the site of at least one excitatory synapse. The volume of the spine has been shown to correlate with the size of the synapse and with the estimated number of functional neurotransmitter receptors present at the synapse. Therefore, the number and size of dendritic spines are considered indicators of the state of synaptic input to a given neuron. Imaging the dendritic spines on a given neuron reveals both the individual states of synapses and their collective distribution on that neuron. Such images may therefore enable us to bridge the gap between systems and molecular approaches to the study of memory. The size of a dendritic spine is typically between 500 and 3000 nanometers. This is close to the size of a wavelength of visible light. The most accurate way to image spines is therefore to use higher-frequency radiation, i.e. electron microscopy. However, this method is extremely time-consuming and yields only a tiny sample size. Our lab has chosen instead to employ 3D confocal microscopy, a technique which allows a larger sample size. The technique yields 3D image stacks of individual dendritic segments. On each dendritic segment, a hundred or more spines appear with blurry outlines. In order to quantify such images, we require a computational algorithm that can segment the images so that individual spines can be counted and measured.
The data is obtained from a confocal microscope at about 240x
magnification.
The 3D dataset (obtained as a set of slices) was then segmented by the team of Prof. Tony Yezzi @ Georgia Tech.
Input
Section of the dataset: 8564 pts.
The "surface scaffold" (i.e., the less significant part of the shock scaffold) gives us back a meshing of the input data. Thus, we can re-build connectivity on the input samples automatically.
Surface mesh
|
|
|
|
|
|
|
|
2002-3
