Search

Background Diffusion tensor imaging (DTI) has successfully been used to investigate ALS-related changes in the central nervous system (CNS). Many studies reported, for example, decreased white matter integrity in the cortico-spinal tract (CST). Consequently, DTI has been proposed as a potential biomarker to support clinical diagnosis. While recent results look promising in principle, the reliability of DTI-based diagnosis seems currently insufficient for practical use. To enhance discriminative performance, one would ideally integrate information from complementary diagnostic modalities during classifier construction. Objective In a recent study, we observed ALS-related thinning of different peripheral nerves based on cross-sectional area (CSA) measurements using high-resolution ultrasound ( Schreiber et al., 2014 ). Here, we combined DTI-based measures of the CST with CSA of the upper limb peripheral nerves to improve imaging-based discrimination between different ALS subtypes and healthy controls. Methods 51 patients with ALS (14 classic, 10 upper motor neuron dominant (UMND), 14 lower motor neuron dominant (LMND), 8 bulbar, and 5 primary lateral sclerosis (PLS)) and 18 healthy controls underwent diffusion-weighted MRI at 3T and peripheral nerve ultrasound in order to determine the fractional anisotropy (FA) in the white matter underneath the motor cortex, posterior limb of the internal capsule, cerebral peduncle, pons and medulla oblongata as well as the CSA of the N. medianus and the N. ulnaris. Statistical comparisons were conducted using analysis of variance (ANOVA). Classifier construction was carried using linear discriminants after dimensionality reduction based on principal component analysis. Separate classifiers were constructed using FA, CSA, and both FA and CSA data, respectively. Results Compared to controls, ALS patients exhibited significant, distally pronounced reductions of the ulnar nerve CSA across all subtypes with the exception of PLS. In addition, DTI data revealed significant differences between ALS subtypes with pronounced impairment of the CST in UMND and PLS and modest impairment in LMND patients, while bulbar ALS is intermediately located within this spectrum. Classification performance was considerably higher exploiting both central and peripheral data compared to using either modality alone. Most notably, UMND and PLS patients could be accurately separated. Conclusion Combining central and peripheral data acquired through DTI and ultrasound, respectively, leads to improved discrimination not only between ALS and healthy controls, but also between ALS subtypes. This speaks to the complementary role of both methods regarding their contribution to classification performance. The potential of combining both imaging methods is particularly evident in the reliable separation of UMND and PLS patients, since early differentiation of these subtypes remains a difficult problem in clinical practice. [ABSTRACT FROM AUTHOR]