Abstract
335. Post-Stroke Cognitive Recovery Can Be Predicted Using Functional and Structural Connectome Estimates
Biological psychiatry (1969), Vol.99(10 Supplement), pp.S243-S244
05/15/2026
DOI: 10.1016/j.biopsych.2026.03.569
Abstract
Background
Stroke frequently results in cognitive impairment, yet the mechanisms of brain recovery remain unclear. While diffusion MRI and resting-state fMRI can measure the brain’s structural (SC) and functional (FC) connectomes, these scans are not routinely used in stroke patients. A scalable alternative is needed to estimate network disruption using only routine clinical imaging (lesion masks).
Methods
We estimated structural (eSC) and functional (eFC) connectomes directly from lesion masks in 132 adults with stroke. The NeMo Tool generated eSCs by removing streamlines intersecting each lesion from a normative tractography atlas. These eSCs were then entered into the Krakencoder, a joint autoencoder model that maps SC to FC, to produce eFC. Observed FC (oFC) was obtained from standard fMRI preprocessing pipelines. Ridge regression was used to predict attention outcomes from eSC, eFC, and oFC.
Results
Both eSC (r = 0.34, pFDR < 1e–3) and eFC (r = 0.36, pFDR = 1e–3) predicted 2-week post-stroke attention scores with accuracy comparable to oFC derived from fMRI data (r = 0.33, pFDR < 1e–3). These findings demonstrate that eSC and eFC capture behaviorally meaningful disruptions to structural and functional brain networks caused by stroke.
Conclusions
This novel pipeline that estimates functional and structural network disruptions from lesion masks demonstrates that it may provide a time-efficient yet behaviorally relevant alternative to advanced MRI in patient populations. This pipeline offers clinicians a practical tool for better understanding how lesions can impact brain networks, and may enable more accurate prognosis and personalized recovery planning for psychiatric impairments following a stroke.
Details
- Title: Subtitle
- 335. Post-Stroke Cognitive Recovery Can Be Predicted Using Functional and Structural Connectome Estimates
- Creators
- Christie Gillies - Weill Cornell MedicineKeith Jamison - Weill Cornell MedicineAaron Boes - University of IowaAmy Kuceyeski - Weill Cornell Medicine
- Resource Type
- Abstract
- Publication Details
- Biological psychiatry (1969), Vol.99(10 Supplement), pp.S243-S244
- DOI
- 10.1016/j.biopsych.2026.03.569
- ISSN
- 0006-3223
- Publisher
- Elsevier Inc
- Language
- English
- Date published
- 05/15/2026
- Academic Unit
- Roy J. Carver Department of Biomedical Engineering; Neurology; Psychiatry; Stead Family Department of Pediatrics; Iowa Neuroscience Institute; Neurology (Pediatrics)
- Record Identifier
- 9985160556602771
Metrics
6 Record Views