TMS Phenotypes of Corticospinal Excitability and Interhemispheric Inhibition in Health and Stroke

Abstract Interhemispheric inhibition and corticospinal excitability are widely used biomarkers of motor cortical function. Yet the normative physiological organization of these measures and their relationship to post-stroke dysfunction remain poorly understood. In this cross-sectional observational study, we analyzed bilateral resting motor threshold (rMT) and ipsilateral silent period (iSP) in a large, harmonized cohort of 142 healthy adults to determine whether healthy motor physiology is organized into physiological phenotypes. Standardized bilateral measures were analyzed using unsupervised clustering. Phenotype structure was further examined using multiple dimensionality-reduction approaches (PR-Isomap, PCA, t-SNE, and UMAP), and supervised classifiers evaluated with repeated stratified 5-fold cross-validation. An independent cohort of 30 individuals with stroke (mean age 65.0 ± 12.8 years; range 40–85; 15 women, 15 men) was then evaluated relative to the fixed healthy-derived reference space without model retraining; 14 participants with complete bilateral measures were eligible for primary four-dimensional projection. A three-phenotype solution was supported in the healthy cohort: High excitability (51/142), Weak IHI (52/142), and Low excitability (39/142). The phenotypes differed in all four physiological measures (all Kruskal–Wallis P < 0.0001) but not in rMT or iSP asymmetry, indicating bilateral physiological differences rather than systematic hemispheric imbalance. Phenotype membership was independent of age, sex, handedness, and musicianship. Phenotype separability was preserved across low-dimensional representations, with mean balanced classification accuracy of approximately 0.90–0.98. In the projectable stroke subgroup, phenotype distribution differed from that of healthy participants (Monte Carlo χ²(2) = 7.00, P = 0.027, Cramér V = 0.21), with High excitability observed in 10/14 stroke participants (71.4%) compared with 51/142 healthy participants (35.9%). High-excitability assignment was more frequent in stroke (odds ratio = 4.46, 95% confidence interval 1.33–14.95; Fisher’s exact P = 0.018). In contrast, nearest-centroid distance did not differ between stroke and healthy participants (median 1.54 versus 1.36; U = 1091, P = 0.550, Cliff’s δ = 0.10). High-excitability enrichment persisted in a sensitivity analysis including 17 stroke participants with at least three of four phenotype-defining measures (11/17, 64.7%; Monte Carlo P = 0.043). These findings demonstrate that healthy motor cortical physiology is organized into reproducible excitability–inhibition phenotypes derived through a multivariate, data-driven framework. Among stroke participants, the principal alteration was redistribution across healthy-derived phenotypes (particularly toward High excitability) rather than global displacement from the healthy phenotype space. Separating phenotype identity from the magnitude of physiological deviation provides a normative framework for interpreting individual post-stroke physiology and may support future biologically informed stratification and precision neuromodulation.