Classification of postural control characteristics during quiet standing in patients with subacute stroke and their association with lesion networks

Standing balance after stroke is crucial for independence and fall prevention, yet quiet-standing postural control and its neural substrates are incompletely understood. We studied 75 patients (mean age 61.08 ± 11.76 years; 59 men) within 2 months of stroke onset. During 30-s eyes-open quiet standing, center-of-pressure (CoP) data were recorded and 52 variables were derived. We applied supervised structured sparse principal component analysis (S3PCA) to extract components and Gaussian mixture clustering to define phenotypes. Admission MRI quantified the percent damage of white-matter tracts using the Lesion Quantification Toolkit. Between-cluster differences in tract damage were assessed using Kruskal–Wallis and Dunn–Holm post hoc tests; within-cluster associations between tract damage and S3PCA scores were assessed using Spearman’s rank correlation. S3PCA yielded four components (antero-posterior frequency distribution, medio-lateral spatial, antero-posterior frequency, and medio-lateral frequency). The three following clusters emerged: (1) higher-frequency profile, (2) lower-frequency/spatial profile, and (3) low medio-lateral frequency with high spatial profile. Compared with cluster 2, cluster 3 showed greater damage in the medial lemniscus and spinothalamic tracts, and tract damage correlated positively with medio-lateral frequency scores within cluster 3. Ascending vestibular/somatosensory pathway damage is linked to phenotype-specific quiet-standing postural control after stroke.