The imaging of complete neural circuits with synaptic resolution is now within reach with automated electron microscopy (EM). The reconstruction of every neuronal arbor and every synapse on every arbor would let us understand the neuroanatomical basis of brain function.

But manually reconstructing every neuron from serial EM is an extremely laborious task: even for a small animal with only 10,000 neurons, the predicted time is about 45 years. Automated techniques must be developed. Over the past 5 years a wealth of machine learning techniques for image segmentation have set the stage for what might be possible.

Our lab has manually reconstructed hundreds of neurons and we are ready to host a PhD project where algorithms are trained from human-generated ground truth for automated reconstruction of thousands of neurons.

You will develop new algorithms that learn the rules of neuronal morphology from human-reconstructed neurons, and apply them for the automated reconstruction of complete neural circuits. You will then analyze these circuits to understand how sensory input converges into higher-order processing centers and intervenes in the generation of motor patterns.

Other parallel projects ongoing in the lab are analyzing the function of neurons that have already been reconstructed, using genetically targeted neuronal activators and inhibitors.

Apply for job