To characterize white matter disease evolution in amyotrophic lateral sclerosis (ALS), using an event-based model (EBM) designed to extract temporal information from cross-sectional data. 

Conventional methods for understanding mechanisms of rapidly progressive neurodegenerative disorders are limited by the subjectivity inherent in the selection of a limited range of measurements, and the need to acquire longitudinal data.  Probabilistic generative models (such as the EBM) are designed to provide a natural staging scheme while also characterizing the uncertainty of the ordering of biomarkers.  Crucially, these models do not require a priori clinical diagnostic information or explicit biomarker threshold criteria.

The EBM characterizes a disease as a series of events, where an event is the change of a biomarker from a ‘healthy’ to a ‘diseased’ state. The model was applied to diffusion MRI data from 154 patients and 130 healthy controls selected from five independent data sets acquired in four different imaging laboratories between 1999 and 2016. The biomarkers modelled were mean fractional anisotropy values of white matter tracts implicated in ALS. The cerebral portion of the corticospinal tracts (CSTs) was divided into 3 segments.  Fine-detailed modelling was performed separately on the CSTs and corpus callosum (CC).

Application of the model to the pooled datasets revealed that the CSTs were involved before other white matter tracts. Distal CST segments were involved earlier than more proximal (i.e. cephalad) segments. The model also revealed early ordering of fractional anisotropy change in the CC and subsequently in long association fibers.

These findings represent data-driven evidence for early involvement of the CSTs and body of the CC in keeping with conventional approaches to image analysis, while providing new evidence to inform directional degeneration of the CSTs. This data-driven model provides new insight into the dynamics of neuronal damage in ALS.