Accuracy of a single, thigh-worn inertial measurement unit for detecting turns during indoor and outdoor community walking in unimpaired adults

Introduction
Turning is a feature of everyday walking estimated to occur during 35% - 45% of daily steps [1]. The instability created during this maneuver can challenge the balance of a range of people such as stroke survivors, amputees and people with Parkinson’s disease [3,4]. Consequently turns have been recognized as a factor in falling [5]. Assessing turning ability is therefore important for mobility screening and to evaluate rehabilitation interventions. While tests such as the time up and go [2] include turning, the artificial and predictable nature of these tests questions their validity for understanding real world mobility. The first step in understanding turns during community walking is an accurate method of measurement
Research Question
Can an inertial measurement unit (IMU) accurately detect turns during indoor and outdoor community walking?
Methods
Ten independent community walkers (height 1.72 ± 0.15 m, weight 68 ± 9.2 Kg, age 21 ± 0.2 years) were recruited. After providing informed consent a single IMU (Paltechnologies, Glasgow, UK), incorporating a magnetometer, was attached to the front thigh of each participant according to manufacturer guidelines. An algorithm (Matlab, MA, USA) based primarily on fluctuations in the magnetometer signal was designed to count the number of turns exceeding 90 degrees, between 60 and 90 degrees and between 30 and 60 degrees. Participants were instructed to walk around an indoor course that included four 90 degree turns and an outdoor course around a university campus that included 12 90 degree turns. A researcher accompanied each participant to ensure compliance with the course. The absolute difference between the 90 degree turns measured by the IMU and actual turns in the course was calculated as a measure of accuracy.
Results
During the indoor course the IMU recorded an average of 4.1 turns (SD 1.2, range 2-6) with an absolute error of 10 turns for the 40 actual turns, or 25%. During the outdoor walks there was an average of 10.7 turns (SD 3.1, range 7-17) with an absolute error of 29 turns for the 120 actual turns, or 24%. (Figure) 5.
Discussion
This study demonstrated the potential for IMUs to count turns during community indoor and outdoor walking. The findings represent a first step towards accurate measurement of turns during everyday community walking. With a 25% detection error the algorithm has potential to improve, an distinctive operationalized definition of a turn would be beneficial. Natural movement variability means that achieving 100% accuracy, in a community setting, is unlikely, however future work will aim to refine the algorithm and test it on mobility impaired individuals.