Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster B in Poster Session B: Tuesday, August 4, 2:00 – 3:45 pm, Kimmel Center, Shorin & Rosenthal Rooms

Spectral structure of rehabilitation experience constrains behavioural recovery after stroke

Noor Sajid¹, Samuel J. Gershman¹; ¹Harvard University

Presenter: Noor Sajid

Patients with similar lesions after stroke can have different outcomes under rehabilitation, indicating that residual capacity alone does not determine behavioural recovery. Instead, it is constrained by whether the post-damage experience spans task-relevant directions needed for relearning function. Here, we formalise recovery as constrained re-optimisation through the surviving network and test whether the spectral structure of rehabilitation experience constrains recovery. We retrain damaged networks under rehabilitation distributions of varying rank, while holding training intensity and learning mechanism constant. Low-rank rehabilitation compresses post-damage activity, reduces effective rank, and limits behavioural recovery, whereas higher-rank experience supports reoccupation of task-relevant directions and improves performance. Across conditions, recovery is predicted by both the rank of the rehabilitation distribution and the effective rank of recovered activity. Post-stroke recovery is therefore constrained not only by structural loss, but also by the spectral breadth of rehabilitation experience available during relearning.

Topic Area: Methods, Tools, Theory & Neural Coding