Motor Network Efficiency in Stroke Patients: Comparing Activation Likelihood Estimation-Based and Whole Brain Parcellations.

Stroke disrupts brain networks, impacting motor function. While whole-brain parcellations offer network-level insights, their relevance to specific motor outcomes may be limited. We compared the ability of two parcellations - Schaefer's 100-parcellation atlas and an Activation Likelihood Estimation-based atlas focused on upper limb motor tasks - to predict upper limb motor outcome in 142 stroke patients. Network efficiency was assessed using Shortest Structural Path Lengths quantified with the Lesion Quantification Toolkit. The Activation Likelihood Estimation (ALE) based parcellation demonstrated superior predictive performance in a k-NN regression model, with higher predictive power, lower error susceptibility, and better model fit compared to the Schaefer parcellation. Our findings suggest that function-specific ALE-based parcellations may provide more accurate and clinically relevant insights into post-stroke upper limb motor outcome by capturing network-level changes specific to the affected function.Clinical Relevance-This suggests that function-specific parcellations, such as ALE-based atlases, may improve the accuracy of predicting motor outcomes in stroke patients.