Predicting Fall Probability Based on a Validated Balance Scale

Alaa Masalha, Nadav Eichler, Shmuel Raz, Adi Toledano-Shubi, Daphna Niv, Ilan Shimshoni, Hagit Hel-Or; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2020, pp. 302-303

Accidental falls are the most frequent injury of old age and have dramatic implications on the individual, family, and the society as a whole. To date, fall prediction estimation is clinical, relying on the expertise of the physiotherapist for performing the diagnosis based on standard scales, such as the highly common and validated Berg Balance Scale (BBS). Unfortunately, the BBS is a time consuming subjective score, prone to variability and inconsistency between examiners. In this study, we developed an objective, computational tool, which automates the BBS fall assessment process and allows easy, efficient and accessible assessment of fall risk. The tool is based on a novel multi depth-camera human motion tracking system integrated with Machine Learning algorithms. The system enables large scale screening of the general public at very little cost while significantly reducing physiotherapist resources. The system was pilot tested in the physiotherapy unit at a major hospital and showed high rates of fall risk predictions as well as correlation with physiotherapists BBS scores on individual BBS motion tasks.