Automatic distinction of upper body motions in the main anatomical planes.

The assessment of spinal mobility and function is gaining clinical importance for the diagnosis and monitoring of low back pain, but its measurement and evaluation remains difficult. As a critical step towards non-supervised assessment of spinal functional, the aim of this study was to assess the efficacy of symmetrical sensors fixed to the sides of the spinal column to distinguish between different upper body movements in the main anatomical planes. 429 healthy volunteers underwent a defined choreography including repeated upper body flexion, extension, lateral bending and axial rotation exercises. The movements were assessed using the Epionics SPINE sensor system. Two pattern recognition models were developed and applied to distinguish between the different movements in a frame-by-frame manner, as well as for whole motion sequences. On average, it was possible to differentiate between different upper body movements with a sensitivity of over 96% for both modelling approaches. The largest type II error was the incorrect identification of extension, possibly due to deviations from the reference standing posture during measurements and small changes in the lordotic angle during extension. The use of two sagittal sensors attached symmetrically to the back therefore seems to allow the distinction of upper body movements in a robust manner, and therefore opens perspectives for the unsupervised recognition of movements and functional activity over extended periods.