Main Menu (Mobile)- Block

Main Menu - Block

Seven Column Connectome (FIB-SEM)

janelia7_blocks-janelia7_fake_breadcrumb | block
node_title | node_title
Seven Column Connectome (FIB-SEM)
node_body | node_body

Motivation

Our most recent (and, as yet, unpublished) reconstruction is focused on seven columns within the medulla, as shown above. The medulla forms hexagonal columnar arrays: one center column with 6 neighbors. Again, one can think of these columns as parallel units in a receptive field. There are two main goals for analyzing 7 adjacent medulla columns:

  • Biologically, adjacent columns when correlated together should uncover the motion detection circuit.  Our seven column dense reconstruction is larger and more comprehensive than our previous ssTEM reconstruction (Takemura'13) and should offer new insights to this critically important circuit. Furthermore, this circuit is the focus of much research including several theoretical models and general physiological studies. With this rich background, we may be able to produce reasonable hypotheses on the circuit dynamics from structural connectivity, which can then guide further focused experimentation.

  • Presumably, these adjacent columns will be very stereotyped and similar in number of neurons, synapses, etc. It provides an opportunity to study stereotypy, understand biological variability, and examine common motifs over similar columns of neuropil. This may also allow one to distinguish biological sources of variability from the variability due to the reconstruction process.

 

Results

  • We were able to identify over 50,000 pre-synapses and over 300,000 post-synaptic connections.
  • We extracted and identified over 500 neuron shapes.
  • We achieved better and faster reconstruction rates with our new FIB-SEM pipeline (compared to ssTEM).  This is a result of better image segmentation and easier tracing from the isotropic resolution.

 

Data Release

  • 7 column connectome and neuron shapes here
  • Segmentation and grayscale from one column of the dataset here