Back Silhouette and Sagittal Vertical Axis Measurement for Hunchback Using Wearable IMU Body Sensor Network

Hunchback is a prevalent condition in modern society, requiring timely treatment. To enable users to monitor their posture at home, this study places five inertial measurement units (IMUs) along the spine on the back surface of the subjects. This paper proposes an algorithm to compute the sagittal plane curvature of the back in a static posture. By integrating measured data with personalized spinal features extracted from a neural network trained on X-ray dataset characteristics, a spinal model is constructed. The accuracy of the algorithm is validated using an optical motion capture system. Compared to ideal values, the spinal model constructed by the algorithm achieves an average Hausdorff distance (HD) of 9.873% and an average mean absolute error (MAE) of 4.830% relative to the total spine length. Additionally, the method yields an average relative error of 14.941% for the sagittal vertical axis. Our results demonstrate that the proposed algorithm can generate a reliable and quantifiable spinal model. Therefore, this method has significant potential for applications in home monitoring and remote medical care.