Topological Clustering of Aphasic Brain Networks

Author(s)

• Jiaying Yi
• Jian Yin
• Rahul Ghosal Ph.D., University of South CarolinaFollow
• Dirk Den Ouden Ph.D., University of South CarolinaFollow
• Julius Fridriksson Ph.D., University of South CarolinaFollow
• Rutvik Desai
• Yuan Wang

Description

Stroke disrupts brain connectivity. A left-hemisphere stroke could lead to post-stroke aphasia (PSA).

Statistical Problems in PSA

• Subtyping: Traditionally based on behavioral scores, ignoring underlying information in the brain.

• Treatment outcome trajectory: Tracking longitudinal changes in the brain over course of treatment.

Modeling Challenge & Solution

• Standard graph-theoretic methods rely on thresholding networks at a single scale.

• Persistent homology (PH) yields multiscale topological features in both signals and networks.

Topological Clustering of Brain Signals & Networks

• State-of-the-art topological clustering methods are highly efficient computationally but require additional transformations for signals.

• We also need a procedure for selecting optimal number of clusters and covariate selection.

Publication Info

2026.

© 2026, Jiaying Yi, Jian Yin , Rahul Ghosal , Dirk B. den Ouden, Julius Fridriksson, Rutvik Desai, & Yuan Wang