Fig. 1: Topological mapping of conversation and progression probabilities.
From: Deep learning of conversation-based ‘filmstrips’ for robust Alzheimer’s disease detection
a, b Example of statements (i.e., speech acts) mapped onto the topological space \((x,y)\): incrementing along x represents the opening or reiteration of a theme, while incrementing along y indicates the elaboration of a sub-theme. c, f Progression probabilities along the x-axis (blue) and y-axis (red) for both recall tasks (Task 1: most pleasant memory; Task 2: least one). Patients with Alzheimer’s disease (solid) display slower decay rates and more dispersed peaks in transition probabilities compared to controls (dashed). d, e, g, h Visualization of aggregated \({(x}_{\max },{y}_{\max })\) topological matrices: in Alzheimer’s patients, a more diffuse distribution in topological positions is observed compared to controls.
