Diagnosis of amyotrophic lateral sclerosis (ALS) can be difficult from clinical symptoms alone. Diffusion tensor imaging (DTI) has been suggested as an adjunct diagnostic method. DTI parameter changes have been repeatedly demonstrated, especially in the corticospinal tract (CST) as the predominantly affected structure. However, a recent meta-analysis reported only a modest discriminatory capability, questioning the value of this method as a confirmatory test in single subjects with suspected ALS. We investigated how methodological differences in CST delineation influence the discriminatory capability.

DTI data were acquired in 13 ALS patients and an age-matched healthy control group. We calculated and compared receiver operation characteristic (ROC) curves of four different analysis methods using either a manual or an atlas-based region of interest (ROI) of the CST in combination with and without tract-based spatial statistics (TBSS).

The analysis method combining atlas-based ROIs with TBSS yielded an area under the curve (AUC) of 0.936 and a sensitivity and specificity of 100 % and 91.67 %. These are the best results among the four analysis methods evaluated: manual ROIs (AUC = 0.846, sensitivity: 69.23, specificity: 91.67), atlas-based ROIs alone (AUC = 0.917, sensitivity: 76.92, specificity: 91.67), manual ROIs in combination with TBSS (AUC = 0.885, sensitivity: 76.92, specificity: 91.67).

Sensitivity and specificity strongly depend on the CST delineation approach. The combination of an atlas-based ROI with TBSS is a promising fully automatic method with improved discriminatory capability compared to other approaches. It could ultimately serve as a confirmatory test in single ALS patients.