Systematic analysis of different low-pass filter cut-off frequencies on lumbar spine kinematics data and the impact on the agreement between accelerometers and an optoelectronic system.

Abstract

A necessary step in the validation of accelerometers for the measurement of spine angles is to determine the levels of agreement with current gold standard methods. However, agreement may be a function of filtering parameters. We aimed to (1) systematically determine the effect of different filter frequency cut-offs on the peak range of motion (ROM) during forward bending as measured by accelerometers and an optoelectronic (OE) system, (2) explore the influence of filtering on agreement between systems, and (3) determine the difference in peak ROM measurement between these systems. Accelerometers and OE sensors were attached at L2, L4, and S1 of 20 asymptomatic female participants for a guided flexion trial. Signals were then iteratively low-pass filtered with cut-off frequencies ranging from 14 Hz to 1 Hz and peak range of motion outcome measures were compared between systems. Peak ROM was minimally affected by filter cut-off frequency for both accelerometer and OE system. The difference in peak ROM between difference cut-off frequencies were maximum 0.66°, median 0.18° and minimum 0.06° for accelerometer derived values and maximum 0.23°, median 0.08° and minimum 0.03° for the OE system. The maximum difference across the filtering frequencies was 0.62° and the largest difference between the two systems (with outliers removed) was 0.82°. Cut-off frequencies ranging from 14 to 1 Hz had little effect of peak lumbar spine ROM during low velocity (6°/s) forward bending, regardless of motion capture method. Filtering cut-off frequency had little effect on the differences between the accelerometer and OE system and similar measurements can be achieved using accelerometers compared to OE systems.