study:
Aesthetic_Neural_Architecture

SCENE FILE:
ARTyficial_signatures

SCENE:
Visual semantics signature

DESCRIPTION:
This scene contains a cortical visualization of the neural signature of the visual semantics dimension of shared aesthetic evaluation (Fig. 3a).

Model weights map based on a linear support vector regression (SVR, C = 1): AATr_predictPC1_FS4_10k_weights.dscalar.nii
Z-scores calculated at each grayordinate based on the bootstrap distribution of the population-level predictive pattern, using 10,000 bootstrap samples (with replacement): AATr_predictPC1_FS4_10k_weights_Zvalue.dscalar.nii

To further evaluate the reliability of the signature weights, we computed the Haufe transformation (SFig. 11a). This transformation converts the population-level pattern (i.e., signature weights, also referred to as the backward model) into an activation pattern (forward model), mapping each grayordinate to the response (fitted values) in the multivariate model: AATr_predictPC1_FS4_10k_weights_Haufe_transformed.dscalar.nii

Regions with significant and consistent model weights were identified after correcting for multiple comparisons (one-sample t-test, FDR q < .05). To further highlight meaningful clusters, we applied minimum size thresholds of 40 mm² for cortical vertices and 80 mm³ for subcortical voxels (equivalent to 10 vertices or voxels). This procedure was applied separately for positive and negative weights:
PC1_10k_pos_k10clusters.dscalar.nii
PC1_10k_neg_k10clusters.dscalar.nii

TAGS:
Species:Human