Transdiagnostic Network Localization of Social, Language, and Motor Symptoms in Patients With Frontotemporal Lobar Degeneration

Abstract

Background and objectives: Frontotemporal lobar degeneration (FTLD) includes different clinical syndromes with distinct patterns of symptoms and neuroanatomical locations of neurodegeneration. However, FTLD is clinically heterogeneous (with overlapping symptoms across several domains) and neuroanatomically heterogeneous (with brain atrophy in different locations in different patients). Traditional methods struggle to fully account for this heterogeneity. In this study, we use a relatively new neuroimaging approach, atrophy network mapping, to localize clinical symptoms in patients with FTLD to specific brain networks transdiagnostically.

Methods: Data were obtained from the Frontotemporal Lobar Degeneration Neuroimaging Initiative and 4-Repeat Tauopathy Neuroimaging Initiative. Inclusion required T1-weighted MRI and a diagnosis of behavioral-variant frontotemporal dementia (bvFTD), semantic-variant primary progressive aphasia (svPPA), nonfluent primary progressive aphasia (nfvPPA), progressive supranuclear palsy Richardson syndrome (PSP-rs), corticobasal syndrome (CBS), or normal cognition. Measures of social cognition (Interpersonal Reactivity Index, Revised Self-Monitoring Scale), language (Boston Naming Test, Animal Fluency), and motor function (Unified Parkinson Disease Rating Scale Part III, PSP Rating Scale) were correlated with neuroimaging measures, including cortical thickness, volume, and atrophy network mapping, a newer method that localizes regions connected to brain atrophy using a functional connectome from cognitively normal persons (n = 1,000).

Results: Fifty-seven patients with bvFTD (age 61.2 ± 6.8 years, 35% female), 41 PSP-rs (age 69.7 ± 7.4 years, 54% female), 39 CBS (age 66.2 ± 6.2 years, 51% female), 37 svPPA (age 63.0 ± 6.0 years, 46% female), and 36 nfvPPA (age 68.3 ± 7.3 years, 53% female) and 135 healthy age-matched controls (age 63.3 ± 7.4 years, 58% female) were included. Compared with atrophy alone, atrophy network mapping showed more consistent neuroimaging results across patients with the same clinical syndrome (Dice index mean 0.68 vs 0.11, paired t = 263.1, df = 4,452, p < 0.001), more strongly explained social cognition (F(1, 84) = 10.2, p = 0.002) and motor symptoms (F(1, 185) = 91.3, p < 0.001) across different syndromes, and showed novel neuroanatomical associations, with the temporal parietal junction relating to social cognition; Wernicke area relating to language symptoms; and association sensorimotor cortex, thalamus, and cerebellum relating to motor symptoms.

Discussion: Atrophy network mapping can improve understanding of brain-behavior relationships in clinically and neuroanatomically heterogeneous disorders such as FTLD.

Figure 1. Atrophy Network Mapping Method

(A) In a sample of age-matched cognitively normal subjects, we generate a normative model for cortical thickness (CT) to estimate the effects of age and sex on CT (measured as linear slopes β_age and β_sex); the resulting residual SD (RSD), derived from the residual r, gives the variability in CT after age and sex correction. (B) Using the normative CT model, for each patient with frontotemporal lobar degeneration (FTLD), we compute a map of w-scores, which corresponds to a z-score for CT adjusted for age and sex. Locations where w ≤−2 are defined to be atrophied. (C) For each FTLD patient's map of atrophy, we compute the voxel-wise connectivity in a normative functional connectome (n = 1,000). The resulting Pearson correlation maps (1 for each subject in connectome) are normalized by Fisher z-transformation and t-tested voxel-wise to yield 1 t-valued atrophy network map per patient with FTLD.

Figure 2. Single-Subject Comparison of Cortical Thickness and Atrophy Network Mapping in FTLD

(A) The proportion of subjects with atrophy (w-score ≤−2) was computed for each FTLD syndrome at each voxel. (B) The proportion of subjects with increased atrophy connectivity (T ≥ 5) was determined at each voxel. bvFTD, behavioral variant frontotemporal dementia; CBS, corticobasal syndrome; FTLD, frontotemporal lobar degeneration; nfvPPA, nonfluent variant primary progressive aphasia; PSP-rs, progressive supranuclear palsy Richardson syndrome; svPPA, semantic variant primary progressive aphasia.

Figure 3. Group Comparison of Cortical Thickness and Atrophy Network Mapping in FTLD

(A) Using a vertexwise linear model with age and sex covariates, we determined locations where patients with different FTLD syndromes had greater atrophy than cognitively normal subjects. (B) A t test was conducted for each FTLD syndrome to determine locations where there was greater atrophy connectivity relative to cognitively normal subjects. bvFTD = behavioral variant frontotemporal dementia; CBS = corticobasal syndrome; FTLD = frontotemporal lobar degeneration; nfvPPA = nonfluent variant primary progressive aphasia; PSP-rs = progressive supranuclear palsy Richardson syndrome; svPPA = semantic variant primary progressive aphasia.

Figure 4. Localization of Clinical Symptoms Using Cortical Thickness and Atrophy Network Mapping in FTLD

(A) Vertexwise cortical thickness was regressed against perspective-taking, naming, and motor abnormality to determine where worsening social-cognitive, language, and motor symptoms would be associated with atrophy. (B) Within patients with FTLD, voxel-wise connectivity to atrophy was regressed against perspective-tasking, naming, and motor abnormality to localize the atrophy network associated with each of these symptoms. (C) A priori ROIs for social cognition, language, and sensorimotor function derived from Neurosynth. (D) For each subject, we determined the connectivity from their map of atrophy to a priori ROIs for symptoms associated with social, language, and motor function. These subject-level connectivity values were regressed against symptom measures. FTLD = frontotemporal lobar degeneration; ROI = region of interest.