Tract-specific analysis improves sensitivity of spinal cord diffusion MRI to cross-sectional and longitudinal changes in amyotrophic lateral sclerosis

Abstract

Amyotrophic lateral sclerosis (ALS) is a late-onset fatal neurodegenerative disease that causes progressive degeneration of motor neurons in the brain and the spinal cord. Corticospinal tract degeneration is a defining feature of ALS. However, there have been very few longitudinal, controlled studies assessing diffusion MRI (dMRI) metrics in different fiber tracts along the spinal cord in general or the corticospinal tract in particular. Here we demonstrate that a tract-specific analysis, with segmentation of ascending and descending tracts in the spinal cord white matter, substantially increases the sensitivity of dMRI to disease-related changes in ALS. Our work also identifies the tracts and spinal levels affected in ALS, supporting electrophysiologic and pathologic evidence of involvement of sensory pathways in ALS. We note changes in diffusion metrics and cord cross-sectional area, with enhanced sensitivity to disease effects through a multimodal analysis, and with strong correlations between these metrics and spinal components of ALSFRS-R.

Fig. 1. Flow diagram of study participation.

ALS participant withdrawals were recorded as being either due to ALS progression or due to reasons unrelated to ALS.

Fig. 2. Cross-sectional differences in FA along the cord.

Graphical representations of cross-sectional differences in FA between ALS (n = 20) and control (n = 20) participants (the one-sided error bars represent SD), and corresponding VCM-corrected p-value heat maps, highlighting spinal levels with significant group differences for whole cord without segmentation, only white matter, and only descending tracts (corticospinal, rubrospinal, reticulospinal, lateral vestibulospinal, and tectospinal). In the p-value heat map, colors red to orange represent p-values < 0.05 and colors yellow to blue represent p-values > 0.05. The sensitivity of the method increases with segmentation. Box plots of the same data are included as Supplementary Information, Supplementary Fig. 1.

Fig. 3. Cross-sectional differences in tract-specific FA along the cord.

Graphical representations of cross-sectional differences in FA between ALS (n = 20) and control (n = 20) participants (the one-sided error bars represent SD), and corresponding VCM-corrected p-value heat maps, highlighting spinal levels with significant group differences for lateral CST and spinal lemniscus. In the p-value heat map, colors red to orange represent p-values < 0.05 and colors yellow to blue represent p-values > 0.05. Box plots of the same data are included as Supplementary Information, Supplementary Fig. 1.

Fig. 4. Cross-sectional differences in CSA along the cord.

Graphical representation of cross-sectional differences in CSA between ALS (n = 20) and control (n = 20) participants (the one-sided error bars represent SD), and corresponding VCM-corrected p-value heat maps, highlighting spinal levels with significant group differences for whole cord, white matter, and gray matter. In the p-value heat map, colors red to orange represent p-values < 0.05 and colors yellow to blue represent p-values > 0.05. Box plots of the same data are included as Supplementary Information, Supplementary Fig. 1.

Fig. 5. Summary of FA and CSA data of whole cord, white/gray matter, and individual tracts.

The dot plots show the mean values (black lines) with 95% confidence intervals (green lines). Sample size n = 20 for ALS and control participants.

Fig. 6. Correlation analysis.

Correlation between ALSFRS-R (Q4-Q12) and (a) mean whole cord FA from C2 to C6, (b) mean whole cord FA at C2, (c) mean whole cord CSA from C2 to C6, (d) mean whole cord CSA at C2, (e) mean white matter CSA at C2, and (f) mean gray matter CSA at C2. Sample size n = 20.

Fig. 7. Longitudinal changes in diffusion metrics.

Longitudinal changes in ALS participants (n = 11), in FA, RD, MD, and AD (averaged across C2–C6) in (a) whole spinal cord, (b) spinal white matter, (c) descending tracts, (d) lateral CST, (e) spinal lemniscus, and (f) posterior columns. The line between initial and one year data connects the means (the numerical values in the text). The box plots show the median (circle with dot), 25th and 75th percentiles (edges of the box), and the range of FA/CSA (the whisker end points) excluding the outliers (individual circles).

Fig. 8. Graphical representation of the VCM-corrected p-value heat maps of group difference in longitudinal changes in RD and MD, in the 1-year follow-up data.

Significant increase, with a size of ~10%, is noted in the group of descending tracts and the lateral CST. In the p-value heat map, colors red to orange represent p-values < 0.05 and colors yellow to blue represent p-values > 0.05.

Fig. 9. MRI data processing and analysis.

a Flow diagram depicting the image processing and feature extraction pipeline. b Representative results of streamlines obtained by tractography in the whole cord and segmentations of the cord into white/gray matter and individual tracts.