Distinct spatiotemporal atrophy patterns in corticobasal syndrome are associated with different underlying pathologies

Abstract

Although the corticobasal syndrome was originally most closely linked with the pathology of corticobasal degeneration, the 2013 Armstrong clinical diagnostic criteria, without the addition of aetiology-specific biomarkers, have limited positive predictive value for identifying corticobasal degeneration pathology in life. Autopsy studies demonstrate considerable pathological heterogeneity in corticobasal syndrome, with corticobasal degeneration pathology accounting for only ∼50% of clinically diagnosed individuals. Individualized disease stage and progression modelling of brain changes in corticobasal syndrome may have utility in predicting this underlying pathological heterogeneity, and in turn improve the design of clinical trials for emerging disease-modifying therapies. The aim of this study was to jointly model the phenotypic and temporal heterogeneity of corticobasal syndrome, to identify unique imaging subtypes based solely on a data-driven assessment of MRI atrophy patterns and then investigate whether these subtypes provide information on the underlying pathology. We applied Subtype and Stage Inference, a machine learning algorithm that identifies groups of individuals with distinct biomarker progression patterns, to a large cohort of 135 individuals with corticobasal syndrome (52 had a pathological or biomarker defined diagnosis) and 252 controls. The model was fit using volumetric features extracted from baseline T1-weighted MRI scans and then used to subtype and stage follow-up scans. The subtypes and stages at follow-up were used to validate the longitudinal consistency of the baseline subtype and stage assignments. We then investigated whether there were differences in associated pathology and clinical phenotype between the subtypes. Subtype and Stage Inference identified at least two distinct and longitudinally stable spatiotemporal subtypes of atrophy progression in corticobasal syndrome; four-repeat-tauopathy confirmed cases were most commonly assigned to the Subcortical subtype (83% of individuals with progressive supranuclear palsy pathology and 75% of individuals with corticobasal-degeneration pathology), whilst those with Alzheimer's pathology were most commonly assigned to the Fronto-parieto-occipital subtype (81% of individuals). Subtype assignment was stable at follow-up (98% of cases), and individuals consistently progressed to higher stages (100% stayed at the same stage or progressed), supporting the model's ability to stage progression. By jointly modelling disease stage and subtype, we provide data-driven evidence for at least two distinct and longitudinally stable spatiotemporal subtypes of atrophy in corticobasal syndrome that are associated with different underlying pathologies. In the absence of sensitive and specific biomarkers, accurately subtyping and staging individuals with corticobasal syndrome at baseline has important implications for screening on entry into clinical trials, as well as for tracking disease progression.

Keywords: biomarkers; corticobasal syndrome; disease progression; machine learning; subtype and stage inference.

Graphical Abstract

Figure 1

Two-subtype model of atrophy progression in CBS identified by subtype and stage inference (SuStaIn). (A) Spatial distribution and severity of atrophy at each SuStaIn stage by Subtype. Each row (Subcortical top, Fronto-parieto-occipital bottom) represents a subtype progression pattern identified by SuStaIn consisting of a set of stages at which brain volumes in CBS cases reach different z-scores relative to controls. Total n = 123 (subcortical n = 56, fronto-parieto-occipital = 67). (B) Assignment of CBS pathology to each SuStaIn subtype. Size of bar (x-axis) represents percentage of cases labelled with that PSP syndrome assigned to that SuStaIn subtype (y-axis). Total n = 123 (PSP = 6, CBD = 12, IDT = 74, AD = 31). PSP = PSP pathology at post-mortem, CBD = at post-mortem, AD = AD pathology at post-mortem or a positive AD biomarker (CSF or amyloid PET) during life. Visualizations in A were generated using the BrainPainter software, modified to include the brainstem segmentations.

Figure 2

Three-subtype model of atrophy progression in CBS identified by subtype and stage inference (SuStaIn). (A) Spatial distribution and severity of atrophy at each SuStaIn stage by Subtype. Each row (subcortical top, fronto-parietal middle and parieto-occipital bottom) represents a subtype progression pattern identified by SuStaIn consisting of a set of stages at which brain volumes in CBS cases reach different z-scores relative to controls. Total n = 122 (subcortical n = 38, fronto-parietal = 56, parieto-occipital = 28). (B) Assignment of CBS pathology to each SuStaIn subtype. Size of bar (x-axis) represents percentage of cases labelled with that PSP syndrome assigned to that SuStaIn subtype (y-axis). Total n = 123 (PSP = 6, CBD = 12, IDT = 73, AD = 31). PSP = PSP pathology at post-mortem, CBD = at post-mortem, AD = AD pathology at post-mortem or a positive AD biomarker (CSF or Amyloid PET) during life.

Figure 3

Stage progression at follow-up visits by SuStaIn subtype. Scatter plots of each subtype for (A) the two-subtype model (B) the three-subtype model showing predicted stage at baseline (x-axis) versus predicted stage at follow-up scan (y-axis) for those subtypable CBS cases with a follow-up scan (n = 103). The area of the circle is weighted by the number of scans at each point, and the colour of the circle represents the time (years) between visits.