Identification of signal-based gait features and blood analytes associated with stroke status and walking speed in mild acute ischemic stroke

Assessment of mild acute ischemic stroke (mAIS) can be difficult, especially in regions where access to neuroimaging is limited. Integrating routine blood analyses with wearable gait assessments may aid in the identification of mAIS and estimation of functional outcomes. In this pilot study, we focus on identifying candidate gait and blood analytes associated with stroke status and functional outcomes in mAIS.

Video, smartphone, and blood data were collected from 22 mAIS patients and nine healthy controls. Ridge-penalized logistic regression nomograms were developed; one to estimate the stroke status using gait, and a second one to estimate walking speed of patients using blood analytes.

Our gait-based nomogram identified mAIS with an area under the curve (AUC) of 0.878, while blood-based nomogram estimated walking speed with an AUC of 0.906 (both optimism-corrected). Identified gait features for stroke status included lower xgyr_energy, xgyr_max, xgyr_std, zgyr_energy and zacc_rms, and higher zgyr_iqr, while higher age, blood urea nitrogen (BUN), and lower estimated glomerular filtration rate (eGFR) and mean corpuscular hemoglobin concentration (MCHC) were associated with slower walking.

We show the feasibility of integrating gait statistics with blood analytes to identify mAIS and estimate walking speed. The nomograms showed strong discrimination and calibration within our small cohort. These findings point to the potential of multimodal signal-based features in stroke detection and rehabilitation planning, particularly in low-resource settings.