Workflows in the CVL Neuroimaging Workbench

The Nectar CVL Neuroimaging sub-project will soon begin work on a neuroimaging workflow. The initial target is a workflow that allows users to perform the complex task of generating a morphometric minimum deformation average of a group of images. This process requires a large amount of pre-processing, checking of input data and is generally only feasible on large datasets via the use of a small compute cluster. For example it takes on the order of a week to generate an average model of a mouse brain using a fairly conservative approach on a small 20 core linux cluster. More advanced versions of this workflow involve permuting the model generation thus extending the processing time.

An example image is given below of what can be achieved with this approach when applied to ex-vivo 18 MRI images of mouse brain at 16.4T. What is evident is that the resulting model while showing the average morphometric signal and intensities doesn not exhibit the vagaries and artifacts visible in the individuals that made up the model. This is especially important when constructing average models of disease or normalcy from a population be they mouse, human or zebrafish images.

Model developed with ex-vivo 18 MRI images of mouse brain at 16.4T

Model developed with ex-vivo 18 MRI images of mouse brain at 16.4T

Given the large size of the images (multiple GB), the Neuro-CVL also automatically includes an installation of the TissueStack web based viewing platform that allows remote viewing and sharing of data. A demo site can be seen running on RDSI here:

Following this workflow we will integrate tools that will allow the registration and matching of ex-vivo optical microscopy data to in-vivo MRI data. The goal of the Neuro-CVL is to both create an easy to use installation of the most commonly used analysis packages and the provision of commonly desired workflows that currently require significant expertise to install and maintain.