Reproduce “Stroke recovery phenotyping through network trajectory approaches and graph neural networks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sanjukta Krishnagopal, Keith Lohse, Robynne Braun
The authors apply trajectory-based clustering (TPC) to NIH Stroke Scale (NIHSS) symptom profiles from 489 patients in an acute stroke trial. They threshold symptoms across five time points, compare patient profiles, and use Louvain community detection to identify groups. TPC found mildly affected, left-motor, and right-motor profiles. A graph neural network classified these internally generated labels, not clinical outcomes.
Key insights
- Network trajectory approaches can effectively identify clinically relevant, distinct stroke recovery subtypes from heterogeneous longitudinal data, offering enhanced interpretability compared to traditional methods.
- Stroke recovery trajectories become largely defined within the first 24 hours post-stroke, highlighting a critical early window for prognostication and potential interventions.
- The predictive dynamics for recovery subtype stratification differ between tPA-treated and placebo patient groups, indicating a sustained influence of tPA on recovery patterns.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.