Each dataset consists of a grayscale image stack and its associated machine segmentation and annotations. To provide good interactive performance for users, image and segmentation planes are divided into tiles at multiple resolutions and stored on a server. NeuTu loads tiles of the desired zoom on demand.
NeuTu relies on DVID to supply data. DVID was created to handle datasets that greatly exceed the memory available on a single computer and, moreover, allows multiple users to simultaneously work on a single dataset. DVID is a distributed, versioned, image-oriented data service designed by Bill Katz to support Janelia Research Campus's brain imaging, analysis and visualization efforts. For more details about DVID and what it offers please visit this site: dvid.io.
Reconstructing a connectome from an EM dataset often requires a large effort of proofreading automatically generated segmentations and synapses. There are two kinds of segmentation errors: false merges and false splits. To extract neuronal bodies from our datasets, proofreaders will interpret the ultrastructure in grayscale images by identify axons, dendrites and synapses in order to edit segmentation to correct falsely merged and over-segmented or falsely split neurons. NeuTu has a suite of tools to identify and correct both types of errors. NeuTu also provides intuitive functions for annotating synapses, including removing false positives, adding missing synapses and correcting synaptic connections.