Table 1.
Descriptive statistics of metric demographic and clinical variables.
Table 2.
Descriptive statistics of categorical demographic and clinical variables.
Table 3.
Penn UMN Score.
Figure 1.
Demonstration of the deformable DTI normalization algorithm used in DTI-TK.
Row (A) shows color maps of DTI data (analogous coronal FA maps) of 4 individuals, with variations in head positioning, shape, and volume. Row (B) shows results after the initial registration to the IXI brain DTI template, with improved positioning. Row (C) shows results after applying optimized affine transforms, which only partly correct for shape and volume variations. Row (D) shows results after applying optimized deformable transforms, with considerably improved alignment of white matter structures and correction for shape and volume differences. These spatial transforms enable tractography, ROI definition, and statistics measurement to be performed in template space.
Figure 2.
ROIs corresponding to the right and left CST, used for template-space measurements.
Figure A depicts the regions-of-interest (ROIs) corresponding to the right (blue) and left (red) corticospinal tracts in template-space, as defined using deterministic tractography. These ROIs were used to measure Mean Diffusivity (MD) and Fractional Anisotropy (FA) in all subjects. As a reference, figure B shows the course of the corticospinal tracts through the brain, taken from http://commons.wikimedia.org/wiki/File:Gray764.png.
Figure 3.
Scatterplots of DTI metrics and the Penn UMN score.
(A): The scatter matrix of MD (unitless) vs. the Penn UMN Score (unitless from 0–32) shows an approximate linear relationship. Superimposed simple regression line is also shown. (B): The scatter matrix of FA (unitless) vs. the Penn UMN Score (unitless from 0–32) shows an approximate linear relationship. Superimposed simple regression line is also shown.
Figure 4.
Added variable plots showing the effect of adding the Penn UMN score to the regression models.
(A): Added variable plot upon adding the Penn UMN Score to the MD regression model. The Penn UMN Score is a significant predictor in the model, with coeff = 0.075, t = 3.1, P = 0.005. (B): Added variable plot upon adding the Penn UMN Score to the FA regression model. The Penn UMN Score is a significant predictor in the model, with coeff = –0.08, t = –3.35, P = 0.003.
Table 4.
Analysis of covariance table for full regression models.
Table 5.
Analysis of covariance table for limited regression models, excluding Clinically Possible ALS.