Automated Brain Volumetry in Patients With Memory Impairment: Comparison of Conventional and Ultrafast 3D T1-Weighted MRI Sequences Using Two Software Packages

Abstract

BACKGROUND. Isotropic 3D T1-weighted imaging has long acquisition times, potentially leading to motion artifact and altered brain volume measurements. Acquisition times may be greatly shortened using an isotropic ultrafast 3D echo-planar imaging (EPI) T1-weighted sequence. OBJECTIVE. The purpose of this article was to compare automated brain volume measurements between conventional 3D T1-weighted imaging and ultrafast 3D EPI T1-weighted imaging. METHODS. This retrospective study included 36 patients (25 women, 11 men; mean age, 68.4 years) with memory impairment who underwent 3-T brain MRI. Examinations included both conventional 3D T1-weighted imaging using inversion recovery gradient-recalled echo sequence (section thickness, 1.0 mm; acquisition time, 3 minutes 4 seconds) and, in patients exhibiting motion, an isotropic ultrafast 3D EPI T1-weighted sequence (section thickness, 1.2 mm; acquisition time, 30 seconds). The 36-patient sample excluded five patients in whom severe motion artifact rendered the conventional sequence of insufficient quality for volume measurements. Automated brain volumetry was performed using NeuroQuant (version 3.0, CorTechs Laboratories) and FreeSurfer (version 7.1.1, Harvard University) software. Volume measurements were compared between sequences for nine regions in each hemisphere. RESULTS. Volumes showed substantial to almost perfect agreement between the two sequences for most regions bilaterally. However, most regions showed significant mean differences between sequences, and Bland-Altman analyses showed consistent systematic biases and wide limits of agreement (LOA). For example, for the left hemisphere using NeuroQuant, volume was significantly greater for the ultrafast sequence in four regions and significantly greater for the conventional sequence in three regions, whereas standardized effect size between sequences was moderate for four regions and large for one region. Using NeuroQuant, mean bias (ultrafast minus conventional) and 95% LOA were greatest in cortical gray matter bilaterally (-50.61 cm(3) [-56.27 cm(3), -44.94 cm(3)] for the left hemisphere; -50.02 cm(3) [-54.88 cm(3), -45.16 cm(3)] for the right hemisphere). The variation between the two sequences was observed in subset analyses of 16 patients with and 20 patients without Alzheimer disease. CONCLUSION. Brain volume measurements show significant differences and systematic biases between the conventional and ultrafast sequences. CLINICAL IMPACT. In patients in whom severe motion artifact precludes use of the conventional sequence, the ultrafast sequence may be useful to enable brain volumetry. However, the current conventional 3D T1-weighted sequence remains preferred in patients who can tolerate the standard examination.