Identifying and predicting gait stability metrics in people with stroke in uneven-surface walking using machine learning

Q: What is the clinical evidence from this study?

Study design: Cross-Sectional. Population: Stroke (n=110). Intervention: Stroke vs. Healthy controls. Primary outcome: Machine learning classifier accuracy in distinguishing people with stroke from healthy controls based on trunk acceleration features during uneven-surface walking (Accuracy >95%).

Yasuhiro Inui; Yusaku Takamura; Yuki Nishi; Shu Morioka

doi:10.1038/s41598-026-35966-9

January 17, 2026Scientific ReportsOpen Access

Identifying and predicting gait stability metrics in people with stroke in uneven-surface walking using machine learning

View Full Paper

Ask AI

Bookmark

Share

Key result

Machine learning using trunk acceleration distinguishes stroke from healthy controls with ~95% accuracy.

Accuracy >95%
n=110

Key Points

To identify trunk acceleration metrics that differentiate people with stroke from healthy controls during uneven-surface walking and determine if even-surface gait parameters can predict them.
Analyzed trunk acceleration data from 71 people with stroke and 39 healthy controls during even- and uneven-surface walking using classification and regression machine learning models.
Evaluated linear and nonlinear acceleration features using model explainability methods, including SHAP values and partial dependence plots.
Machine learning classifiers differentiated people with stroke from healthy controls with an accuracy of over 95%, identifying vertical root mean square (RMS_VT), anterior-posterior sample entropy (SampEn_AP), and harmonic ratio (HR_AP) as key discriminative features.
Gait speed on uneven surfaces was highly predictable from even-surface walking parameters in linear (R² = 0.903) and nonlinear models (R² = 0.860), with an even-surface speed < 0.8 m/s predicting greater speed reduction and elevated RMS_VT.
Uneven-surface metrics were predicted by even-surface features and kinematics, including RMS_VT (RF, R² = 0.630), HR_AP (SVR, R² = 0.607), and SampEn_AP (Elastic Net, R² = 0.393), which exhibited a nonlinear relationship with ankle dorsiflexion.

Study Design

Type

Cross-Sectional (n=110)

Multicenter

No

PICO

P

Population

110 adults (71 with stroke and 39 healthy controls) capable of independent walking underwent cross-sectional analysis of trunk acceleration during even- and uneven-surface walking tasks.

E

Exposure / Comparator

Stroke vs Healthy controls

O

Primary Outcome

Machine learning classifier accuracy in distinguishing people with stroke from healthy controls based on trunk acceleration features during uneven-surface walking — Accuracy >95%

Main Result

Effect estimate: Accuracy >95%

Limitations

Single-center study with a modest number of healthy controls (n=39) limiting statistical power and generalizability.
Data augmentation techniques were used but do not substitute for real participant data.
Feature selection was conducted prior to train-test splitting, potentially introducing a risk of information leakage.
Predictive accuracy of SampEn_AP was limited, likely due to inherently noisy entropy measures and relatively short recording lengths (ten gait cycles).
Gait parameters like SampEn were derived only from the paretic side, not fully capturing bilateral asymmetry.

Implication

Analysis identifies gait stability metrics in stroke survivors, suggesting potential for personalized rehabilitation.

Abstract

People with stroke (PwS) face increased fall risk on uneven surfaces; however, gait stability under such conditions remains unexplored. This study used machine learning (ML) to identify acceleration features distinguishing PwS from healthy controls (HC) during uneven-surface walking and to predict them from even-surface gait parameters. Trunk acceleration data from 71 PwS and 39 HC were analyzed using classification and regression models. The ML classifiers achieved an accuracy of over 95%. The key discriminative features included the vertical root mean square (RMS_VT), anterior-posterior sample entropy (SampEn_AP), and harmonic ratio (HR_AP). In PwS, even-surface gait speed < 0.8 m/s predicted reduced speed and higher RMS_VT on uneven surfaces. SampEn_AP and HR_AP were influenced by ankle kinematics and their even-surface values, respectively, showing nonlinear associations. These findings support the use of wearable sensor data and interpretable ML to assess gait stability and adaptability, facilitating development of digital biomarkers for personalized stroke rehabilitation aimed at improving outdoor mobility.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Inui et al. (2026) conducted a cross-sectional in Stroke (n=110). Stroke vs. Healthy controls was evaluated on Machine learning classifier accuracy in distinguishing people with stroke from healthy controls based on trunk acceleration features during uneven-surface walking (Accuracy >95%). Machine learning classifiers using trunk acceleration features (RMS_VT, SampEn_AP, and HR_AP) distinguished people with stroke from healthy controls during uneven-surface walking with over 95% accuracy.

synapsesocial.com/papers/696f1ac19e64f732b51eefe4 https://doi.org/10.1038/s41598-026-35966-9

View Full Paper

Ask AI

Bookmark

Share

Identifying and predicting gait stability metrics in people with stroke in uneven-surface walking using machine learning

Key Points

Study Design

PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Identifying and predicting gait stability metrics in people with stroke in uneven-surface walking using machine learning

Key Points

Study Design

PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading