Versions Compared


  • This line was added.
  • This line was removed.
  • Formatting was changed.
Comment: Migrated to Confluence 4.0

Overview is a software package optimized to perform image registration on mouse brains. Registration may be defined as the fitting of two or more images to an average image. Differences between the average image and the individual images allows us to calculate mean differences between 2 or more groups of mice. Some examples of groups of mice which may be compared are (1) Mice homozygous for mutation A, heterozygous for mutation A, and wild type mice (2) Mice homozygous for mutation A and mice homozygous for mutation B (where it remains to be determined whether gene A and B perform the same function) (3) Mice that are given a specific treatment versus those that are not. Differences between groups may also be examined by determining the precise location of differences and the amount of difference.

An assumption in registration is that every point in a brain can be mapped to a point in another brain. Without this, registration fails. This is why we cannot perform registration on brain images that are drastically different from each other. There are a number of problems that can creep up which may result in a failure to register. These are detailed in the 'Troubleshooting' section.

Special Note: Although the parameters of the software are specifically set for mouse brains, may also be used for embryos (tail may cause problems) , whole mouse body, human brain, etc.. It is not suitable for organs for which homology is indeterminable (e.g. vasculature).

6 Basic Steps

There are 6 steps carried out in A basic overview is provided below:

1. Rigid alignment, referred to as 6 parameter alignment, consists of rotating and sliding an image to fit a model. The model may be a pre-existing model or a model created through bootstrapping. Using a pre-existing model carries the advantages of saving computation time, saving disk space, and also results in data that is in the desired orientation. In bootstrapping, a model is created out of your dataset. Once 6 parameter alignment is complete, all brain images are now in the same space.

2. The first level of improvement is now carried out. Affine alignment, also known as 12 parameter alignment, consists not only of rotating and translating an image, but also scaling and shearing an image to produce an average. It is carried out in a pairwise manner.

3. Non-linear alignment is then carried out in which the remaining differences are dealt with. Initially, the atlas model from the 12 parameter alignment is used as a target. Then, progressively (from non-linear atlas model 1 to non-linear atlas model 6), improvements are carried out until there is a final average model. The transformation begins with a blurry image that is gradually clarified as the linear transformation proceeds. Settings for this portion of the script are configurable for images with higher or lower resolution.

4. Add segmentation: A segmented atlas brain (in which individual brain structures have been defined) is used to add segmentation to the average brain.

5. Backpropagate segmentation: Segmentation added to the average brain is now transferred to the individual brains.

6. The intensities are stored in an output folder along with the following variables (which are calculated for each brain image): displacement, magnitude, scaled jacobians, and jacobians. These variables are useful in analyzing the data.

The end result is an average brain and deformations for each individual brain image. For descriptions of the variables in #6 above, please see 'What to do with MICe-build-model output'.

Output Folders

Below is a summary of the data contained in the output folders. 'PROJECT_NAME' will differ depending upon the name provided by the user when running MICe-build-model.

1. PROJECT_NAME_lsq6: Contains the average brain following 6 parameter alignment.

2. PROJECT_NAME_lsq12: Contains the average brain following 12 parameter alignment.

3. PROJECT_NAME_nlin: Contains the final average brain following non-linear alignment.

4. PROJECT_NAME_resampled_atlas: Contains the resampled atlas brain. This is the final average brain with each structure defined and labeled using a previously labeled atlas brain.

5. PROJECT_NAME_processed: Contains deformations for each individual brain.

6. PROJECT_NAME_registration_accuracy: Contains 3 folders which contain brain images that are 2, 3, and 4 standard deviations away from the average brain in terms of intensities and number of deformations.This information is useful in determining the cause of failure when registration cannot be completed. The image below is an example of how data is displayed in this folder. The brain image has a total of 2425 voxels highlighted in either green or blue. The green displays voxels at which intensity is at least 4 standard deviations away from the average. The blue displays voxels at which the number of deformations are at least 4 standard deviations away from the average.