Synaptic Cleft Segmentation in Non-Isotropic Volume Electron Microscopy of the Complete Drosophila Brain

Neural circuit reconstruction at single synapse resolution is increasingly\nrecognized as crucially important to decipher the function of biological\nnervous systems. Volume electron microscopy in serial transmission or scanning\nmode has been demonstrated to provide the necessary resolution to segment or\ntrace all neurites and to annotate all synaptic connections.\n Automatic annotation of synaptic connections has been done successfully in\nnear isotropic electron microscopy of vertebrate model organisms. Results on\nnon-isotropic data in insect models, however, are not yet on par with human\nannotation.\n We designed a new 3D-U-Net architecture to optimally represent isotropic\nfields of view in non-isotropic data. We used regression on a signed distance\ntransform of manually annotated synaptic clefts of the CREMI challenge dataset\nto train this model and observed significant improvement over the state of the\nart.\n We developed open source software for optimized parallel prediction on very\nlarge volumetric datasets and applied our model to predict synaptic clefts in a\n50 tera-voxels dataset of the complete Drosophila brain. Our model generalizes\nwell to areas far away from where training data was available.\n