Near-real-time analysis of one-dimensional biomechanical continuum data using random field theory lookup tables

Statistical parametric mapping (SPM) has been used to analyse biomechanical data including force, kinematic and EMG trajectories, but common SPM software implementations require 20 ms for probability calculations, which may be too slow for real-time applications. This study aimed to improve the efficiency of SPM probability calculations and to obtain sub-ms speeds, thereby making these analysis methods appropriate for near-real time analysis and feedback. To this end we constructed linear lookup tables (LLUT) for probabilities, as a function of degrees of freedom and temporal smoothness, and interpolated as appropriate for the smoothness characteristics of particular biomechanics datasets. This LLUT approach resulted in probability calculations of approximately 0.1 ms. These results suggest that SPM may be suitable for near-real time applications including rehabilitation robotics and internet-based analyses.