Fat-suppressed T2-weighted fluid attenuated inversion recovery MR images allow efficient skull stripping in brain tumors using a brain extraction tool

Abstract

Objective: To explore the efficacy of fat-saturated T2-weighted fluid attenuated inversion recovery MRI (fs-FLAIR) in skull stripping for the treatment of brain tumors.

Methods: MR images of brain tumors (gliomas, N=46; meningioma, N=51) were retrospectively collected, in which MRI protocols included contrast-enhanced T1-weighted MRI (T1C), fs-FLAIR or non-fs FLAIR MRI (nfs-FLAIR), T1-weighted MRI (T1), and T2-weighted fast spin echo MRI (T2). Skull stripping was implemented using the Brain Extraction Tool (BET) and evaluated with the Dice similarity coefficient as a comparison to manually segmented brain areas. To test the differences in Dice coefficients across different MR modalities, paired t tests and independent t test were utilized. Spearman’s correlation analysis was used to determine the correlations between Dice coefficients and scanning parameters.

Results: No significant correlations were observed between Dice coefficients and scanning factors influencing image contrast in the fs- FLAIR images of the two datasets, whereas significant correlations were observed with T1, T1C, T2, and nfs-FLAIR images. In gliomas, fs-FLAIR has the best skull-stripping performance, and the Dice coefficients were generally greater than 0.80 (maximum of 0.90). In contrast, most Dice coefficients were less than 0.8 in other sequences. All Dice coefficients of the fs-FLAIR images were significantly greater than those of T1, T2, and T1C images (p < 0.0001). Similar skull-stripping performances were observed in fs-FLAIR images of meningiomas and gliomas. Moreover, compared with nfs-FLAIR, fs-FLAIR resulted in higher Dice coefficients, with a maximum Dice coefficient of 0.87.

Conclusion: Fs-FLAIR allows fast and accurate skull stripping for brain tumors, and has the potential to aid in the development of intelligent diagnosis methods for these tumors.