Alternative thresholding methods for fMRI data optimized for surgical planning

Abstract

Current methods for thresholding functional magnetic resonance imaging (fMRI) maps are based on the well-known hypothesis-test framework, optimal for addressing novel theoretical claims. However, these methods as typically practiced have a strong bias toward protecting the null hypothesis, and thus may not provide an optimal balance between specificity and sensitivity in forming activation maps for surgical planning. Maps based on hypothesis-test thresholds are also highly sensitive to sample size and signal-to-noise ratio, whereas many clinical applications require methods that are robust to these effects. We propose a new thresholding method, optimized for surgical planning, based on normalized amplitude thresholding. We show that this method produces activation maps that are more reproducible and more predictive of postoperative cognitive outcome than maps produced with current standard thresholding methods.

Keywords: Functional MRI; Preoperative mapping; Statistical testing; Thresholding.

Figure 1

Example from a single subject of the three thresholding methods compared in this paper: A) p-value B) AMPLE and C) n-signal.

Figure 2

Second example from another representative single subject of the three thresholding methods compared in this paper: A) p-value B) AMPLE and C) n-signal.

Figure 3

Examples of average percent brain activation. This figure is not meant to represent any particular thresholding method, but to orient the reader to the novel metric of percent brain activation.

Figure 4

Increase in R² after adding LIs derived from each thresholding method to multiple regression model for each memory test

Equation 1

AMPLE method, assuming equal variance among voxels. The left side of the equation simplifies to percent signal of peak voxel

Equation 2

AMPLE method, not assuming equal variance among voxels. The equation simplifies to percent signal, weighted by the ratio of standard deviation differences among the peak and current voxels.

Equation 3

Dice coefficient for quantifying the overlap in two data sets, X and Y. |X| refers to the number of activated voxels in the first split-half dataset, |Y| refers to the activated voxels in the second split-half dataset. |X ∩ Y| refers to the number of voxels that are activated in the same location in both split-half datasets.

Equation 4

Laterality index (LI) used to predict memory outcomes based on distribution of fMRI activation. L refers to the number of voxels above threshold in the left-sided ROI and R refers to the number of voxels on the right.