single neuron imaging 2D morphology 3D morphology VNC imaging tracing & identification mapping & visualization

STEP 1: Segmentation/Tracing

 

approach #1: from 2D segmentation to 3D reconstruction

More details can be found in [Tsechpenakis et al, IEEE EMBC 2011]


approach #2: 3D shape-driven co-segmentation

 

More details can be found in [Tsechpenakis et al, IEEE TBME 2012]



approach #3: Neuron Tracing

 

 

Data can be downloaded here.
More details can be found in [Gulyanon et al, ISBI 2016]

 

STEP 2: Classification



We use a latent Condition Random Field, a variation of Hidden Conditional Random Fields, for part-based neuron classi cation. The objective is the estimation of the motor neuron subtype y, given the unknown compartments t (soma, axon, dendrites) and their corresponding morphology features r.



Open Challenges

new morphology classes from mutations



morphology dynamics: change of morphology and neuron identity



neuronal circuit dynamics