CCB Test Data Archive

Method

Brain masks were created for all cases, using Brain Extraction Tool (BET, FSL). All masks were touched up by neuroanatomists, and reliability measures were performed to ensure that all cortical CSF was included and that all meninges were removed. After the masks were touched up, the brain was extracted, new masks were created for the cerebellum, and the cerebellum was extracted. Cortical object models were created from the skull stripped brains using MNI's Cortical Surface Extraction. Sulcal lines were then traced on these models using MNI�s Display program. The sulci were labeled using the numerical identifiers specified at the LONI website (see http://www.loni.ucla.edu/NCRR) and included:

• Central Sulcus (#1)
• Superior Temporal Sulcus Main Body (#2a)
• Postcentral Sulcus (#3)
• Middle Frontal Sulcus (#5)
• Intraparietal Sulcus (#15)
• Sylvian Fissure (#16)
• Precentral Sulcus (#19)

It also is worth noting that all masks used for extraction of various anatomies have been preserved, such that for example, for the CEREBELLUM directory, a CEREBELLUM_MASK directory can also be found, which contains the masks used for the extraction.

Although it is our hope to have more sub-cortical structures extracted for later versions, for now, only the children have had the Putamen delineated and extracted in both Native and ICBM space.

The XFM directory, is found under the NATIVE directory and contains the transformation matrices needed to go from Native space to ICBM space. Similarly, transformations from ICBM to NATIVE space are found under the ICBM directory.

The SKULL_STRIPPED_NO_CEREB_TISSUE_CLASSIFIED directories will contain, first, two sub-directories, NO_BFC_CORRECTION and BFC_CORRECTION. Within the BFC_CORRECTED directory, all inputs into the various Tissue-Classified algorithms have been corrected for scanner inhomogeneities, using the Bias Field Corrector program (BFC, Brain Suite). Under the NO_BFC_CORRECTION, the files were not corrected for scanner inhomogeneities. The algorithms used for tissue classifications are:

• Partial Volume Classifier (PVC, BRIAN SUITE)
• Partial Volume Extraction (PVE)
• FAST10 (FSL, FAST Tissue classifier, 10 iterations)

From these algorithms, probability maps were created for each case, and for each tissue type, with values ranging from 1 to 3, 1 representing the fact that only one of the algorithms picked the voxel as a particular tissue type, and 3 denoting that all three algorithms picked that voxel as a particular tissue type.