Skip secondary menu

Web-Based System for Labeling of 3D Electron Microscopy Connectome Data

Amauche Emenari Photo

Boston University (Massachusetts)

The Connectome Project aims to provide an unparalleled compilation of neural data, an interface to graphically navigate this data and the opportunity to achieve never before realized conclusions about the brain. The long-term vision of this project is to generate map of the neural connections between individual neurons in order understand how the organization of cells contributes to the function of the nervous system.

As part of this project, Dr. Briggman’s Circuit Dynamics and Connectivity Unit lab in conjunction with the Neuroscience Bioinformatics Program of NINDS is developing a Web-based application and a mobile client for data visualization and 3D interaction to allow users to submitneuronal labels of 3D Electron Microscopy Connectome data. The application provides data visualization and 3D interaction to crowd-source tracing neural membranes in electron microscopy images.  The primary function of this application is to solicit human judgments to create membership functions for segmentations of objects (neurons, mitochondria, ER, etc.).  Users view the raw image data and join (through a “color by click” or contour interface) constituent contours that belong to the same neuron.

During the NINDS summer internship, I have contributed to the server side development through writing code to perform image processing, tracing data processing, and storage and retrieval of user rendered data.I was also responsible for enhancing the frontend experience bylinking to live data stored in the database, adding additional tracing toolsand updating the 3D rendering. Eventually, the crowd-sourced data provided through the Web-Application will be supplied to an artificial neural network as training data.The network will use the data to build a template for accurately detectingthe presence of neural membranes in electron microscope images. After processing hundreds of terabytes of images, the system is expected to generate a detailed model of neural connections in the brain.

Last updated November 26, 2013