Which Volume Reports should be used?
Three different types of volume reports are provided in this analysis as described above.
The raw volumes (mm^3) in the native patient coordinates are the most intuitive numbers. However, the brain size has a large amount of cross-subject variability and, thus, the volumes of the brain structures have the large standard deviations. This may negatively impact the sensitivity of the analysis. For example, the figure below shows the absolute volume of the frontal lobe of 199 subjects as a function of age (left graph). The standard deviation is 12.6% of the average size. From this figure, it is not clear how much of this variability reflects the variability specific to the frontal lobe. For example, there is a tendency of smaller volumes in older subjects, but it could be due to the loss of the entire brain volume, not specific to the frontal lobe.
If the brain size is normalized (right graph), the size of the frontal lobe is approximately 10% of the entire brain and the standard deviation reduced to 6.1%, more clearly revealing the age-dependent loss of the frontal lobe. The relative volumes, however, have disadvantages too. First, it depends on the normalization factor (denominator). For example, the denominator could be tissue volume including or not including the ventricles.
These approaches could be insensitive to the shrinkage of the entire brain. If we are interested in the global brain atrophy, the denominator should include the CSF outside the brain (or intracranial volume), although we found the CSF volumes are often unstable depending the signal from bone marrow. Using the tables of the raw volumes, users can calculate various types of the relative volumes.
The "corrected_stats" table is one of these options we prefer to use, in which the denominator is the total volume of tissues + ventricles + sulci. Another approach is to use the volumes after the linear normalization to the MNI space. After this operation, the total brain volumes of all data become almost the same, which has the effect very similar to the relative volume measurements. In practice, there is no answer to "which approach is correct". Users are advised to look at all three approaches.
