Brain Morphometrics Correlations With Age Among 350 Participants Imaged With Both 3T and 7T MRI: 7T Improves Statistical Power and Reduces Required Sample Size

Abstract

Magnetic resonance imaging (MRI) at 7T has a superior signal-to-noise ratio to 3T but also presents higher signal inhomogeneities and geometric distortions. A key knowledge gap is to robustly investigate the sensitivity and accuracy of 3T and 7T MRI in assessing brain morphometrics. This study aims to (a) aggregate a large number of paired 3T and 7T scans to evaluate their differences in quantitative brain morphological assessment using a widely available brain segmentation tool, FreeSurfer, as well as to (b) examine the impact of normalization methods for subject variability and smaller sample sizes on data analysis. A total of 401 healthy participants aged 29-68 were imaged at both 3T and 7T. Structural T1-weighted magnetization-prepared rapid gradient-echo (MPRAGE) images were processed and segmented using FreeSurfer. To account for head size variability, the brain volumes underwent intracranial volume (ICV) correction using the Residual (regression model) and Proportional (simple division to ICV) methods. The resulting volumes and thicknesses were correlated with age using Pearson's correlation and false discovery rate correction. The correlations were also calculated in increasing sample size from three to the whole sample to estimate the sample size required to detect aging-related brain variation. Three hundred and fifty subjects (208 females) passed the image quality control, with 51 subjects excluded due to excessive motion artifacts on 3T, 7T, or both. 7T MRI showed an overall stronger correlation between morphometrics and age and a larger number of significantly correlated brain volumes and cortical thicknesses. While the ICV is consistent between both field strengths, the Residual normalization method shows markedly higher correlation with age for 3T when compared with the Proportional normalization method. The 7T results are consistent regardless of the normalization method used. In a large cohort of healthy participants with paired 3T and 7T scans, we compared the statistical performance in assessing age-related brain morphological changes. Our study reaffirmed the inverse correlation between brain volumes and cortical thicknesses and age and highlighted varying correlations in different brain regions and normalization methods at 3T and 7T. 7T imaging significantly improves statistical power and thus reduces the required sample size.

Keywords: 3T; 7T; aging; brain morphometrics; magnetic resonance imaging.

FIGURE 1

The 7T has a stronger inverse correlation of total cortical gray matter volume, total subcortical gray matter volume, total white matter volume, and mean cortical thickness with age. Brain morphometric correlations with age using 350 pairs of 3T and 7T MPRAGE scans, including the raw volumes (no correction) and corrected for ICV using either Residuals from regression model (Residual method) or division by ICV (Proportional method).

FIGURE 2

In the subset of the data (N = 47) with comparable resolutions (0.8 mm iso and 0.75 mm iso, 6:35 and 5:02 min at 3T and 7T both with acceleration = 2), 7T images exhibited a higher gray matter to white matter contrast to noise ratio (CNR). The signal‐to‐noise ratio (SNR) in the sample image was calculated by dividing the signal intensity by the standard deviation of the background noise in an artifact free region in‐slice. Note: MPRAGE acquisitions are not considered most suitable for SNR calculations and this calculation does not account for the characteristic non‐Gaussian noise from MRI images.

FIGURE 3

The 7T shows more significant regions in correlation of total gray matter volume or mean cortical thickness with age. Cortical regions corrected for ICV using Residual method and sex and found significant after FDR correction are shown with their respective Pearson's correlation coefficient (positive correlation, red; insignificant, gray; negative correlation, blue). Vessel‐affected regions were removed. Cortical thicknesses were only corrected for sex.

FIGURE 4

The 7T reduces required sample size in all regions, cortical volumes, subcortical volumes, and cortical thickness. The number of significant regions in raw volumes (no correction) and corrected for ICV using both the Residual and Proportional methods observed with increasing sample size significantly differed between 3T and 7T.

FIGURE 5

The 7T‐derived ICV is consistent with that derived from 3T but more accurate in regional volumes. Comparison between the ICV value calculated at 3T and 7T as well as the effect of different ICV correction (Residual and Proportional) methods. Ideal correction should result in no correlation between total cortical volume and ICV. Dashed lines represent 95% confidence intervals. For the Residual method, both correlations showed no significant non‐zero slope. For the Proportional method, 7 T data showed no significant non‐zero slope (p = 0.17) while the 3T data showed a non‐zero slope (p < 0.0001).

FIGURE 6

Mean cortical volume annual rate of change measured at 0.32% for both 3T and 7T. Cortical volumes were corrected for ICV using the Residual method, and the median age of the population used for this analysis is 52 years.