Lesion-Derived Estimated Functional and Structural Connectomes Predict Post-Stroke Recovery Outcomes

Accurate prediction of stroke recovery outcomes has long been challenging, making it difficult for clinicians to forecast individual patient progress. Advanced connectome MRI techniques offer potential prognostic insights into affected brain networks; however, their impracticality has led to the development of lesion-based estimates of network disruptions instead.

This study introduces a novel pipeline that utilizes advanced connectome analysis tools to estimate structural disconnection (eSDC) and functional connectivity (eFC) directly from lesion masks of 96 stroke patients, with the aim of predicting motor, language, attention, and memory outcomes at various time points following stroke.

The researchers employ a Network Modification (NeMo) Tool, which calculates structural disconnectivity (eSDC) using white matter tractography, and a Krakencoder, a joint autoencoder that estimates functional connectomes (eFC) based on the output of the NeMo Tool. The models constructed from eSDC and eFC demonstrate superior performance in predicting almost all outcomes, particularly motor outcomes, compared to models utilizing patients' fMRI-based observed functional connectivity (oFC) or models based solely on lesion information and covariates.

The estimated connectome measures hold several key advantages. Firstly, they retain meaningful inter-individual variability, unlike observed functional connectomes. Secondly, they reflect population-level stroke-related changes in functional connectomes, providing a broader context for individual predictions. Lastly, the models generated from these estimated connectome measures showcase feature importances that align with expected networks of relevance to the given outcome, confirming their biological significance.

Feature importance analyses further reveal distinct predictive signatures across different modalities and outcome domains, indicating that both eSDC and eFC capture non-redundant, biologically meaningful information about post-stroke network disruption. These findings establish lesion-based estimates of post-stroke structural and functional connectomes as a practical, information-rich alternative to advanced imaging techniques in patients.

This approach enables a more detailed exploration of the neural circuits affected by stroke and their role in impairment and recovery, ultimately paving the way for more efficient, accessible, and personalized approaches to post-stroke prognosis and intervention.