Poster Abstracts

BACKGROUND AND AIM:Fall injuries are responsible for significant disability among elderly[1]. Exergames based on full body movements provide a challenging opportunity for training and evaluation of postural control[2], but affordable sensor technology and algorithms for assessment of whole body movement patterns are scarce. Aim: to evaluate the suitability of Kinect, an affordable motion capture system (MCS), for evaluation of balance ability during exergaming. To this end we compared Kinect with a marker based (MB) MCS by 1) detecting common features in whole body MCS data, 2) identifying the contribution of individual anatomical landmarks (ALs) to movement patterns, and 3) by computing outcome measures representing balance ability. We hypothesize that the MB MCS outperforms Kinect, but that evaluation of movement patterns by Kinect is feasible. METHODS:Twenty healthy adults (age 36.95±16.64) performed five different 1-min task-embedded mediolateral weight-shift exergame trials while 2D position data of 10 ALs in the frontal plane was obtained using a 12 camera Vicon V8 system and Kinect. Game conditions included: neutral, lifting a leg and increased game speed, sway amplitude (SA) and sway frequency (SF). Principal Component Analysis (PCA)[3] was performed on normalized Vicon and Kinect data. Common features were identified by projecting data on the three main principal components (PCs). Vicon and Kinect data were combined in a PCA and the contribution of each AL to the three main PCs was quantified by eigenvector analysis. Finally, outcome measures representing balance ability, including SA, SF, RMS and index of harmonicity (IH)[4], were computed from identified common features. RESULTS:Variance explained (VE) by common features for Vicon and Kinect is 94.3±3.8% and 93.6±3.3% respectively, indicating that both systems can cover most of the signal features within three PCs. The contribution of ALs to the movement patterns showed differences in VE ranging from 5% for trunk ALs up to 25% for extremities in all conditions. Differences ranging 4-52% were observed for SA, RMS and IH, while SF difference was within 1-6% range (fig1). CONCLUSION:For the first time PCA was used to compare movement patterns during exergaming. Kinect and Vicon identify common features in MCS data with similar accuracy. However, differences in the contribution of ALs to PCs are observed between devices, thereby affecting accuracy of balance evaluation. These limitations should be taken into account when adopting Kinect for whole body motion capture for balance assessment during exergaming. 1. Sterling DA, et al. J. Trauma 2001, 50:116-9 2. Van Diest M, et al J. Neuroeng. Rehabil. 2013, 10:101 3. Daffertshofer A, et al. Clin. Biomech. 2004, 19:415-28 4. Lamoth CJC, et al. Gait Posture 2002, 16:101-14 Fig1. Balance outcome measures computed from Vicon and Kinect PCA projections along the three main PCs. Vicon and Kinect are indicated by black and grey bars respectively