next up previous contents
Next: Segmentation Up: Use of the DODECANTS Previous: Preparation of the Data   Contents

Registration

The DODECANTS tool is designed to use a single MR image volume to define a standard coordinate system, allowing comparisons of regional CSF volume measurements across groups of subjects. An in-house standard has been used in the work conducted within the TINA core group, and is available at

http://www.tina-vision.net/tina-main/tarballs/DODECANTS/atrophy_standard_brain.tar.gz

If users of the DODECANTS tool wish to use the same standard, then this data set should be downloaded and extracted, and the path to the norm.img file entered into the data list (without the .img extension). The file is in ANALYZE format. More widely used coordinate systems (MNI Brainweb, Taliarach etc.) could be substituted, but users should be aware that this will require preparation of the masks used in the segmentation stage. The masks required for the TINA standard are contained in the tarball listed above, in the file norm_bincut. These should be left in the same directory as the standard brain image.

The TINA MI registration algorithm uses simplex optimisation: since this is a local optimisation routine, a rough alignment of the source data set to the target is required as a starting point. The rough alignment should be prepared in advance, and the results stored in a .tair file, following the instructions given in the TINA User's Guide. It need only bring the images into a very rough alignment, and so can be prepared by manual rotation, translation and scaling. The rough alignment will be applied to all data sets listed in the data list: if the data sets are themselves very poorly aligned (e.g. consist of a mixture of data sets with axial and coronal slices) then they should be split into several data lists, and analysed separately.

Once the rough .tair file and data list have been prepared, the registration stage can be started. Run a tinaTool as described above, using the oct_test macro to set up the required windows. Then enter the path to the directory into the ``Directory" field of the DODECANTS tool, and the name of the data list file (with any extension) into the ``Data List" field. Then press the ``Bulk Coreg" button in the DODECANTS tool. The tool will then run through all data sets listed in the data list, registering them to the standard brain image, and writing the results to .tair files in the same directory as the original data. For example, if the data set is called alice.img, the registration result will be called alice.tair_mi.

Since the registration uses a local optimisation, it may fail on a small number of data sets. Therefore, prior to segmentation, the registration results should be checked. With the path and name of the data list entered into the relevant fields of the DODECANTS tool, press the ``CC start" button. This will initiate the coregistration checker. Then press the ``CC next button". This will load the standard brain, the first data set in the data list, and the registration result into the Coreg tool, and display them in the three Coreg Tv's. The result can be visually inspected. Pressing the ``CC next" button again will advance to the next data set in the data list. When the end of the data list is reached, terminate the coregistration checker by pressing the ``CC end" button.

If the registration has failed on any of the data sets, (i.e. the data set is obviously not aligned to the standard brain), it can be repeated using the Coreg tool whilst still running the coregistration checker, by loading in the rough .tair file from the Coreg tool and performing some manual alignment followed by another attempt at automatic alignment. Remember to remove the previous .tair_mi file and write out the updated .tair file, using the naming convention described above, once a satisfactory alignment has been achieved. See the TINA User's Guide for further instructions on the use of the Coreg Tool.


next up previous contents
Next: Segmentation Up: Use of the DODECANTS Previous: Preparation of the Data   Contents
root 2017-11-24