Skip to site navigation

Reconstruction of neural circuits from electron microscope data

What are we doing?

We are exploring methods for automating the reconstruction of neural circuits from high-resolution electron microscope data. This is an extremely challenging image processing task and fully automating an error-free reconstruction is likely to be unachievable in the immediate future. Therefore our work focusses on reducing the amount of manual effort required for reconstruction, increasing the size of the circuits that can be feasibly reconstructed.

Why are we doing it?

The ultimate goal of neuroscience is to understand the working of the brain. In order to understand how the brain functions, we must understand both how individual neurons function and how they are connected. The "cortical column" is often considered as the classic repeated functional unit in the cerebral cortex, with different columns performing the same computation with different inputs. In the barrel cortex of mice and rats each column processes the input from a different whisker, while in the visual cortex each column processes input from a different part of the retina (corresponding to different locations in visual space). As columns within each cortical area perform the same function it is likely that they will share very similar circuitry and that understanding a single column will therefore contribute greatly to our understanding of the processing that occurs in that cortical area.

Such cortical columns are typically 1mm³ in volume, and recent advances in electron microscopy have brought us closer to being able to image this volume at sufficiently high resolution. However, our reconstruction capability lags far behind. With the best computer-based manual tracing methods, reconstructing a cortical column would take hundreds to thousands of person-years. Reducing the manual effort required for reconstruction is therefore crucial in order to make the reconstruction of a cortical column feasible.

Reconstructing neural networks from electron microscope images

Reconstructing neural networks from electron microscope images