Dynamic balance diagnostics with smart insoles

Introduction: The real-time calculation of the centre of pressure (COP) during walking is of paramount importance for balance assessment and fall prevention. The aim of this study is to derive a balance index (BI) from the COP and its cyclogram and to test how well the BI works in distinguishing between diabetic and healthy volunteers.

Methods: This study involved 15 healthy volunteers and 14 people with diabetes. During walking, the COP was measured using Pedar-X insoles. The position of the COP was calculated at 10 locations: heel strike (E), toe-off (A), the beginning and end of the double support phase (D,B), and intersection points (C) on the COP cyclogram. The standard deviation of each COP cluster was calculated in the X and Y directions and multiplied (area of COP movement within ±1 standard deviation). The BI was calculated as the sum of the 10 areas. Furthermore, the ratio R=BC/AB was calculated from the average data of each cluster. The sensitivity and specificity of the logarithmically transformed BI and R were used to calculate the ROC, AUC, and the classification threshold.

Results and Discussion: LogBI was normally distributed in the diabetic (D) and healthy (H) volunteers, with significantly different (p<0.0001) averages of 1.686±0.228 and 2.070±0.282, respectively. The AUC was 0.861. At an optimal classification threshold of 1.9, the false positive H-data (identified as D) and the false negative D-data (identified as H) were approximately 20% each. The averages of logR were significantly different (p<0.0001): 1.054±0.312 and 0.143±0.306 for H and D, respectively. The AUC was 0.992. At an optimal threshold of 0.5, the false positive H-data and the false negative D-data were both about 4%.

Conclusions: The position of the COP within the cyclogram (logR) performed better than the standard deviation of the COP (logBI).