Multi-Linear Regressor for Static Posturography Estimation Through an Instrumented Cane

Measuring static postural sway outside of the clinic could provide clinicians with long-term, continuous data on patient balance, offering a comprehensive view beyond infrequent in-clinic assessments. This paper presents a novel method to quantify balance ability through a regression algorithm that predicts postural sway velocity using only motion and force sensors. Data is acquired through sensors onboard an instrumented cane. The prediction algorithm's validity was demonstrated in a study of eight young unimpaired subjects and eight adults over 65. The subjects' balance was challenged with different stance widths and sight conditions while using an instrumented cane. In the younger subject cohort, balance was further challenged through an unstable platform. Together, these conditions allowed for variation of the tasks' difficulty levels and thus the range of measured sway velocity. Across subjects, sway velocity was demonstrated to be highly predictable (Younger Subjects $R^{2}=0.73$, Older Subjects $R^{2}= 0.47$) using just the sensors onboard the instrumented cane. In particular, hand motion was shown to be important in predicting sway velocity. We also demonstrated the use of data features to estimate Romberg quotients of the older participants, suggesting the method's potential to track proprioceptive function over time (Correlation $\mathbf{r}=0.82$). This method offers a promising approach to continuous patient monitoring and could provide a long-term, quantitative assessment of balance ability.