UCLA

Purpose

To compare "standardization," "Gaussian normalization," and "Z-score normalization" intensity transformation techniques in dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI) estimates of cerebral blood volume (CBV) in human gliomas. DSC-MRI is a well-established biomarker for CBV in brain tumors; however, DSC-MRI estimates of CBV are semiquantitative. The use of image intensity transformation algorithms provides a mechanism for obtaining quantitatively similar CBV maps with the same intensity scaling.

Materials and methods

The coefficient of variance (CV) in normal-appearing white matter and relative contrast between tumor regions and normal tissue was compared between the three CBV transformations across five different MR scanners in 96 patients with gliomas.

Results

The results suggest all normalization techniques improved variability and relative tumor contrast of CBV measurements compared with nonnormalized CBV maps. The results suggest Gaussian normalization of CBV maps provided slightly lower CV in normal white matter and provided slightly higher tumor contrast for glioblastomas (WHO grade IV) compared with other techniques.

Conclusion

The results suggest Gaussian normalization of leakage-corrected CBV maps may be the best choice for image intensity correction for use in large-scale, multicenter clinical trials where MR scanners and protocols vary widely due to ease of implementation, lowest variability, and highest tumor to normal tissue contrast.