Estimated intracranial volume from FreeSurfer is biased by total brain volume

Abstract

Background: Estimated intracranial volume (eTIV) from FreeSurfer is not segmentation-based but calculated from the alignment of the input magnetic resonance (MR) images to the MNI305 brain atlas, an approach that could lead to a bias by total brain volume. If eTIV is unbiased, variance beyond that explained by intracranial volume should be random. Our null hypothesis was that no correlation would remain between eTIV and total brain volume when controlling for intracranial volume. Methods: eTIV and total brain volume for 62 participants were calculated on 1.5-T, T1-weighted MR images using FreeSurfer (version 6.0.0). Manual delineations of the intracranial volume were also made for the same images. To evaluate the null hypothesis, the partial correlation between eTIV and total brain volume was calculated when controlling for intracranial volume. Results: The partial correlation between eTIV and total brain volume when controlling for intracranial volume was 0.355 (p = 0.026). The null hypothesis was rejected. Conclusion: eTIV from FreeSurfer is biased by total brain volume.