Volume estimation of the brain, white matter, and gray matter using FreeSurfer and FSL: consistency between methods
Humberto Ávila, Vanessa Raulino Silva, Daniel Souza Ferreira Magalhães · Research on Biomedical Engineering · 2019
Magnetic resonance imaging (MRI) usage is increasing during last years and has become the method of choice for the investigation of neuroanatomy in vivo. At either end of the MRI analysis spectrum, there are manual and automated approaches; manual approaches are user-dependent and time-consuming, but are considered to be the gold standard of MR image analysis techniques. We compared the brain, white matter (WM), and gray matter (GM) volumes of automated segmentation in order to evaluate the process of segmentation. We analyzed 59 health subject images between 31 and 82 years old from an image bank, obtaining volumes by two of the most used software for brain processing, FSL and FreeSurfer. We also proposed a new method to improve brain segmentation volume with FSL. A difference of 35%, 46%, and 9% for brain, white matter, and gray matter volumes, respectively, was shown. The volumes obtained with FSL and FreeSurfer were significantly correlated ( p < 0.05) with correlation coefficients of 0.39, 0.79, and 0.52 for the brain, WM, and GM. After the proposed method for FSL brain volume correction (FSLcorr), the correlation coefficient improved from 0.39 to 0.77. Volumes obtained with FreeSurfer were significantly smaller than those with FSL ( p < 0.05) and using FSLcorr method, it was possible to improve FSL measurements.