Hierarchical spectral clustering reveals brain size and shape changes in asymptomatic carriers of C9orf72

Abstract

Traditional methods for detecting asymptomatic brain changes in neurodegenerative diseases such as Alzheimer's disease or frontotemporal degeneration typically evaluate changes in volume at a predefined level of granularity, e.g. voxel-wise or in a priori defined cortical volumes of interest. Here, we apply a method based on hierarchical spectral clustering, a graph-based partitioning technique. Our method uses multiple levels of segmentation for detecting changes in a data-driven, unbiased, comprehensive manner within a standard statistical framework. Furthermore, spectral clustering allows for detection of changes in shape along with changes in size. We performed tensor-based morphometry to detect changes in the Genetic Frontotemporal dementia Initiative asymptomatic and symptomatic frontotemporal degeneration mutation carriers using hierarchical spectral clustering and compared the outcome to that obtained with a more conventional voxel-wise tensor- and voxel-based morphometric analysis. In the symptomatic groups, the hierarchical spectral clustering-based method yielded results that were largely in line with those obtained with the voxel-wise approach. In asymptomatic C9orf72 expansion carriers, spectral clustering detected changes in size in medial temporal cortex that voxel-wise methods could only detect in the symptomatic phase. Furthermore, in the asymptomatic and the symptomatic phases, the spectral clustering approach detected changes in shape in the premotor cortex in C9orf72. In summary, the present study shows the merit of hierarchical spectral clustering for data-driven segmentation and detection of structural changes in the symptomatic and asymptomatic stages of monogenic frontotemporal degeneration.

Keywords: brain segmentation; genetic frontotemporal dementia; shape; size; structural MRI; tensor-based morphometry.

Graphical abstract

Figure 1

Schematic of workflow. (A) Global-to-local segmentation with visualization of the 1st to 6th level of the hierarchical segmentation with circular dendrograms (3th to 8th level is visualized in Supplementary Figure 2 for completeness and 3D images were uploaded as Supplementary data). Intra-segment voxels are randomly coloured according to their position in hierarchical diagrams. For each segment, the transversal plane with the highest number of intra-segment voxels is visualized. (B) Statistical analysis of size and shape components per segment, calculated respectively by means of a Student’s t test comparing the average Jacobians and CCA comparing the principal components (PCs), and link to the circular dendrogram summarizing significance levels across segments (−log₁₀P-value).

Figure 2

Asymptomatic carriers C9orf72 versus non-carriers. Global-to-local segment results for A size and B shape and their respective dendrograms. Asterisk indicates FDR-adjusted significance (dep) P = 0.0007, −log P = 3.17, results below the FDR-adjusted significance threshold are not illustrated. Nodes can be linked to their spatial coverage via Fig. 1A. Results from other levels are shown in Supplementary Figure 3. (C) VBM and TBM analysis: asterisk indicates FDR-adjusted significance (dep) P = 0.0003, t = 3.46. Note that the by convention, only atrophy is visualized so widening of the ventricles cannot be assessed from the univariate results.

Figure 3

Symptomatic carriers C9orf72 versus non-carriers. Global-to-local segment results for A size and B shape and their respective dendrograms. Asterisk indicates FDR-adjusted significance (dep) P = 0.0007, −log P = 3.17, results below the FDR-adjusted significance threshold are not illustrated. Nodes can be linked to their spatial coverage via Fig. 1A. Results from other levels are shown in Supplementary Figure 4. (C) VBM and TBM analysis: asterisk indicates FDR-adjusted significance (dep) P = 0.0003, t = 3.46.

Figure 4

Symptomatic carriers GRN versus non-carriers. Global-to-local segment results for A size and B shape and their respective dendrograms. Asterisk indicates FDR-adjusted significance (dep) P = 0.0007, −log P = 3.17, results below the FDR-adjusted significance threshold are not illustrated. Nodes can be linked to their spatial coverage via Fig. 1A. Results from other levels are shown in Supplementary Figure 5. (C) VBM and TBM analysis: asterisk indicates FDR-adjusted significance (dep) P = 0.0003, t = 3.46.

Figure 5

Symptomatic carriers MAPT versus non-carriers. Global-to-local segment results for A size and B shape and their respective dendrograms. Asterisk indicates FDR-adjusted significance (dep) P = 0.0007, −log P = 3.17, results below the FDR-adjusted significance threshold are not illustrated. Nodes can be linked to their spatial coverage via Fig. 1A. Results from other levels are shown in Supplementary Figure 6. (C) VBM and TBM analysis: asterisk indicates FDR-adjusted significance (dep) P = 0.0003, t = 3.46.

Figure 6

Stratification per disease phenotype and size versus shape results. (A) Jacobians weighted by means of the thresholded maps for size and shape changes per genetic group. (*) indicates between-phenotype differences within the symptomatic groups (Post hoc comparison Tukey–Kramer P < 0.05), note that we did not include the non-carriers and asymptomatic carriers in this statistical comparison. Plots were generated using the Robust Statistic Toolbox. (B) Pair-wise Dice coefficients between the thresholded maps for (a)symptomatic C9orf72 carriers (respectively a, s) and symptomatic GRN and MAPT carriers (s). Note that for size × size and shape × shape comparisons, the Dice coefficient on the diagonal is one (identical) but that the comparison between size and shape results in an asymmetrical matrix with values of <1 on the diagonal.