IMU-Based quantitative assessment of stroke from gait
Yiou Sun, Zhenhua Song, Lifen Mo, Binbin Li, Fengyan Liang, Ming Yin, Dong Wang
Abstract
Gait impairment, which is commonly observed in stroke survivors, underscores the imperative of rehabilitating walking function. Wearable inertial measurement units (IMUs) can capture gait parameters in stroke patients, becoming a promising tool for objective and quantifiable gait assessment. Optimal sensor placement for stroke assessment that involves optimal combinations of features (kinematics) is required to improve stroke assessment accuracy while reducing the number of sensors to achieve a convenient IMU scheme for both clinical and home assessment; however, previous studies lack comprehensive discussions on the optimal sensor placement and features. To obtain an optimal sensor placement for stroke assessment, this study investigated the impact of IMU placement on stroke assessment based on gait data and clinical scores of 16 stroke patients. Stepwise regression was performed to select the kinematics most correlated with stroke assessment (lower limb part of Fugl-Meyer assessment). Sensors at different locations were combined into 28 sensor groups and their stroke assessment was compared. First, the reduced number of gait features does not significantly impact the stroke assessment. Second, the selected gait parameters by stepwise regression are found all from sensors at the hip and bilateral thighs. Last, a three-sensor scheme–sensors at the hip and bilateral thighs was suggested, which achieved a high accuracy with an adjusted R 2 = 0.999, MAE = 0.07, and RMSE = 0.08. Further, the prediction error is zero if the predicted lower limb Fugl-Meyer scales are rounded to the nearest integer. These findings offer a convenient IMU solution for quantitatively assessing stroke patients. Therefore, the IMU-based stroke assessment provides a promising complementary tool for clinical assessment and home rehabilitation of stroke patients.