Iterative Model Building (MBM 2.0)(MBM.py)

We are currently putting the finishing touches on a new, more flexible version of MICe-build-model, currently dubbed MBM 2.0. At some point in summer 2014, it will become the de facto standard used at mice for our standard, iterative, group-wise registration procedure. 

MBM 2.0 is being written in python using the pydpiper framework.

Introduction

The initial class design and documentation for MBM 2.0 can be found here.

How to Use MBM 2.0

• Using an Initial Model
• recipes on how to use MBM.py

Quality Control on your pipeline

Throughout an MBM pipeline (as of version 1.13.1 and up) a number of verification images are created that are meant to be used to get a quick sense of how well your pipeline is moving along.

The very beginning - prior to the lsq6 registration

The first image that is created shows a slice through each of your input files and through the target file for the registration. The target file can be a file from the initial model, or one specified through --lsq6-target or the first input file if you are running a bootstrap model.

Here is a example of an pipeline with 3 input files using an initial model for the lsq6 registration

The LSQ6 registration

The previous image is important to look at in case your initial model, input files and parameters for the registration do not make sense. If the target (initial model) is in a very different orientation than your input files, your lsq6 registration might not work, and you'll also know why that is. The next quality control image that is produced is a slice through your lsq6 average file and each of the input files. If you see that things terribly failed here, you should stop your pipeline and adjust the parameters for the lsq6 stage.

Here is an example of a pipeline with 3 input files. The first file is the overall average, the next three are the individual files after lsq6 alignment: