Background Measurements of Huntington’s disease (HD) progression, such as clinical test scores and medical imaging, can provide powerful markers for clinical applications. However, individual data can be confounded by inter-subject variability, measurement noise and hidden variables, such as disease stage. Disease progression modelling uses probabilistic methods to untangle confounding effects and hence learn patterns of disease-related changes directly from data.

Aims Here we present recent developments in disease progression modelling, which we apply to i) uncover insights into HD, and ii) provide new staging and prognosis utility for clinical applications in HD.

Methods We perform analyses on both cross-sectional and longitudinal data. In the cross-sectional analysis, we use the Event-Based Model (EBM) to reveal the sequence of regional brain volume and clinical marker changes from post-processed structural MRI (sMRI) data from the TRACK-HD, PREDICT-HD, and IMAGE-HD studies. In the longitudinal analysis, we use the Temporal Event-Based Model (TEBM) to reveal a timed sequence of regional brain volume changes from post-processed sMRI data from the TRACK-HD study. We also use both models to learn individual-level information, such as disease stage and survival probability.

Results In the cross-sectional analysis, we find that EBM uncovers a broadly consistent order of events across all three studies; that EBM stage reflects clinical stage; and that EBM stage is related to age and genetic burden. In the longitudinal analysis, we find a new sequence of regional brain volume trajectory events that occur over a period of approximately 16 years; and that TEBM stage is predictive of clinical progression.

Conclusions We have shown the application of two disease progression models to extract information from cross-sectional and longitudinal datasets. These methods can reveal otherwise hidden information, which could be used in clinical staging and prognosis.