Normalization of white matter intensity on T1-weighted images of patients with acquired central nervous system demyelination

Abstract

Background: Intensity variation between magnetic resonance images (MRI) hinders comparison of tissue intensity distributions in multicenter MRI studies of brain diseases. The available intensity normalization techniques generally work well in healthy subjects but not in the presence of pathologies that affect tissue intensity. One such disease is multiple sclerosis (MS), which is associated with lesions that prominently affect white matter (WM).

Objective: To develop a T1-weighted (T1w) image intensity normalization method that is independent of WM intensity, and to quantitatively evaluate its performance.

Methods and subjects: We calculated median intensity of grey matter and intraconal orbital fat on T1w images. Using these two reference tissue intensities we calculated a linear normalization function and applied this to the T1w images to produce normalized T1w (NT1) images. We assessed performance of our normalization method for interscanner, interprotocol, and longitudinal normalization variability, and calculated the utility of the normalization method for lesion analyses in clinical trials.

Results: Statistical modeling showed marked decreases in T1w intensity differences after normalization (P < .0001).

Conclusions: We developed a WM-independent T1w MRI normalization method and tested its performance. This method is suitable for longitudinal multicenter clinical studies for the assessment of the recovery or progression of disease affecting WM.

Keywords: Intensity normalization; clinical trials; magnetic resonance imaging; multiple sclerosis.