Post-acquisition processing confounds in brain volumetric quantification of white matter hyperintensities

Abstract

Background: Disparate research sites using identical or near-identical magnetic resonance imaging (MRI) acquisition techniques often produce results that demonstrate significant variability regarding volumetric quantification of white matter hyperintensities (WMH) in the aging population. The sources of such variability have not previously been fully explored.

New method: 3D FLAIR sequences from a group of randomly selected aged subjects were analyzed to identify sources-of-variability in post-acquisition processing that can be problematic when comparing WMH volumetric data across disparate sites. The methods developed focused on standardizing post-acquisition protocol processing methods to develop a protocol with less than 0.5% inter-rater variance.

Results: A series of experiments using standard MRI acquisition sequences explored post-acquisition sources-of-variability in the quantification of WMH volumetric data. Sources-of-variability included: the choice of image center, software suite and version, thresholding selection, and manual editing procedures (when used). Controlling for the identified sources-of-variability led to a protocol with less than 0.5% variability between independent raters in post-acquisition WMH volumetric quantification.

Comparison with existing method(s): Post-acquisition processing techniques can introduce an average variance approaching 15% in WMH volume quantification despite identical scan acquisitions. Understanding and controlling for such sources-of-variability can reduce post-acquisition quantitative image processing variance to less than 0.5%.

Discussion: Considerations of potential sources-of-variability in MRI volume quantification techniques and reduction in such variability is imperative to allow for reliable cross-site and cross-study comparisons.

Keywords: Cerebrovascular disease; Sources of variability; Volumetric analysis; White matter hyperintensity.

Figure 1:

Flow chart summarizing the use of the discovery dataset (n=21) that examined distinct sources of variability inherent in white matter hyperintensity volumetric quantification (WMH-VQ) processing techniques.

Figure 2:

Common hyperintensity signal artifacts in the white matter hyperintensity (WMH) mask include: Gray matter signals (GM), panels A, and B, (arrowhead); Lateral sulcus and pineal gland, panels A, and B, (rectangle); Voxels in between and inside the ventricles, panels A, B, C, and D, (narrow arrow); Voxels in cerebellum panels E and F (circle); Voxels in the pons and lower slices panels G and H (large arrow).

Figure 3:

Example of a case where WMH masks differ based on SPM versions used. A: is the original T2 FLAIR image. B: WMH mask using MATLAB 2015 and SPM8. It shows an overestimate volume comparing to the FLAIR image and C which is the WMH mask that quantified using MATLAB 2015 and SPM12.

Figure 4.

Several examples of cases that highlight the effect of the center of gravity (CoG) on bone extraction method using BET-FSL tools. Panel A: demonstrates optimal bone extraction with almost clean brain tissue. Panel B and C show non-brain tissue remaining (narrow arrows) due to choosing an alternate CoG. Panel D demonstrates a loss of a portion of GM due to the non-tissue extraction process as a result of choosing an alternate CoG.

Figure 5.

Regression curve for WMH volumes before and after editing (Panel A and B, respectively, (n = 50)). Panel C, the mean value of WMH volume for both raters before and after editing (n = 50). R2 = 0.999, Standard error estimation before editing 118.7 and after editing 68.1.( p < 0.001).