Topographical data analysis to identify high-density clusters in stroke patients undergoing post-acute rehabilitation

Area of Science:

• Neuroscience
• Data Science
• Rehabilitation Medicine

Background:

• Stroke rehabilitation outcomes, including motor and cognitive function recovery, exhibit significant patient variability.
• Identifying distinct patient subgroups is crucial for understanding and optimizing recovery trajectories.

Purpose of the Study:

• To identify data-driven stroke patient subgroups using Topological Data Analysis (TDA).
• To compare TDA with K-means clustering for subgroup identification.
• To assess demographic and clinical differences among identified subgroups.

Main Methods:

• Secondary analysis of clinical, functional, and demographic data from 339 stroke patients.
• Application of TDA, persistent homology, and density clustering to identify recovery subgroups.
• Statistical comparison of inter-group differences using ANOVA, Kruskal-Wallis, and chi-squared tests.

Main Results:

• TDA identified three recovery subgroups: poor (n=34), intermediate (n=88), and good (n=15) recoverers.
• Significant differences were found across subgroups in amantadine use, stroke risk factors, creatinine levels, length of stay, discharge destination, and FIM/MRFS scores (all p < 0.05).
• TDA demonstrated effectiveness in identifying clinically relevant patient stratification.

Conclusions:

• Stroke recovery is influenced by both initial functional status and the presence of stroke risk factors.
• TDA offers a powerful approach for stratifying stroke patients into distinct recovery trajectories.
• Future research incorporating omics data with TDA can accelerate personalized treatment plans for stroke survivors.