Objective
To use an inertial measurement unit (IMU) sensor instead of a 3D optical motion capture system to improve the accuracy of the PCA-BP (principal component analysis-back propagation) model and increase the model prediction task (hip, knee and ankle moment on the affected side of the stairs in stroke patients).
Methods
Inertial Measurement Unit (IMU) and Kistler force plates were used to collect kinematic and kinetic data of 30 stroke patients when walking, ascending and descending stairs. Opensim was used to calculate the hip, knee and ankle moment of stroke patients, and PCA was used to screen the initial variables with a cumulative contribution rate of 99%, and the standard root mean squared error (NRMSE), root mean squared error (RMSE) and mean absolute percentage error were used error, MAPE) and mean absolute error (MAE) and R2 were used as the evaluation indexes of the PCA-BP model. Pearson correlation coefficient (PCC) was used to evaluate the consistency between the calculated and predicted moments.
Results
PCA data showed that the trunk, pelvis, hip, knee and ankle joints on the affected side had a significant effect on the moment of the hip, knee and ankle on the affected side in the x, y, and z axes (x, y, and z were the sagittal axis, coronal, and vertical axes, respectively). The NRMSE was 4.14%~5.26%, the RMSE was 0.132 ~ 0.194, the MAPE was 1.6%~2.9%, the MAE was 0.108 ~ 0.147, and the R2 was ≥ 0.99.
Conclusion
The established PCA-BP model can more accurately predict the hip, knee and ankle moment on the affected side of stroke patients, and the model can also accurately predict the hip, knee and ankle moment on the affected side when patients go up and down stairs, which significantly shortens the measurement time. In addition, in the gait analysis of stroke patients, the IMU sensor can replace the traditional 3D optical motion capture system, so that the patient is not limited by the laboratory environment, and the sports scientists and therapists are more convenient and concise in clinical treatment research.