Comparative reliability analysis of publicly available software packages for automatic intracranial volume estimation

Intracranial volume is an important measure in brain research often used as a correction factor in inter subject studies. The current study investigates the resulting outcome in terms of the type of software used for automatically estimating ICV measure. Five groups of 70 subjects are considered, including adult controls (AC) (n=11), adult with dementia (AD) (n=11), pediatric controls (PC) (n=18) and two groups of pediatric epilepsy subjects (PE1.5 and PE3) (n=30) using 1.5 T and 3T scanners, respectively. Reference measurements were calculated for each subject by manually tracing intracranial cavity without sub-sampling. Four publicly available software packages (AFNI, Freesurfer, FSL, and SPM) were examined in their ability to automatically estimate ICV across the five groups. Linear regression analyses suggest that reference measurement discrepancy could be explained best by SPM [R² = 0.67;p < 0.01] for the AC group, Freesurfer [R² = 0.46; p = 0.02] for the AD group, AFNI [R²=0.97; p < 0.01] for the PC group and FSL [R² = 0.6; p = 0.1] for the PE1.5 and [R² = 0.6; p < 0.01] for PE3 groups. The study demonstrates that the choice of the automated software for ICV estimation is dependent on the population under consideration and whether the software used is atlas-based or not.