Inertial Sensor Based Fall Risk Assessment andSystem Development for the Community-dwelling Older People

Abstract

Along with the global trend of population ageing, falls present a substantial public health problem among elderly people over the age of 65. The objective of this research was to develop a low-cost, portable and inertial sensors based tool for assessing falls risk in the older people. To achieve this goal, three stages of studies have were conducted. The first stage aimed to design a test protocol based on the human balance system for assessing the risk of falling. The test protocol consisted of seven main tests, i.e., sensory integration test, limits of stability test, sit-to-stand five times test, timed up and go test, motor function test, reaction test, and short falls efficacy scale international. Another study was also conducted to examine the effectiveness of developed reaction test APP (application) on assessing cognitive function and fall risk in elderly people. The second stage aimed to conduct large-scale experimental studies to examine the effectiveness of the test protocols on classifying fallers and non-fallers and identifying the underlying causes of high risk of falling. The final stage aimed to develop an inertial sensor-based fall-risk assessment prototype system to assess fall risk for future use with elderly people.

In terms of classifying fallers and non-fallers, we found that the fallers had worse performances than non-fallers on physiological, psychological and integrated functions of the human balance system. Among all fall-risk measures, ten most important measures were the information processing speed in the reaction test, short falls efficacy scale international score in fear of falling test, power density spectral (PSD) of acceleration medio-lateral (ML) for the vision system, angular velocity anterior-posterior (AP) for the vision system, PSD of angular velocity AP for postural stability, sit-stand jerk in the sit-to-stand five times test, PSD of angular velocity AP for the vision system, sit-stand duration in sit-to-stand five times test, angular velocity AP in timed up and go test, and maximal turning angular velocity in timed up and go test. Furthermore, six typical models were developed to classify fallers and non-fallers based on significant measures, including logistic regression (LR), linear discriminant analysis (LDA), classification and regression tree (CART), boosted tree (BT), random forest (RF), and support vector machine radial basic function (SVMRBF) models. The results indicated that the BT, RF, and SVMRBF models had excellent accuracy (>85%). The CART model had good accuracy (>75%), but the LDA and LR models had relatively low accuracies of about 70%. In order to identify the underlying causes of high fall risks, the CART-PA method, which integrated the CART model and profile assessment method, was proposed to identify the factors of high risks of falling. The CART-PA method could generate reinforced results from these two methods, which not only identifies the main factors but also possible factors of high fall risks. Therefore, the CART-PA method could be a useful complementary tool for identifying underlying causes of high fall risks. Fall assessment prototype system included two parts, i.e., hardware and software. The hardware contained five wireless inertial sensors and one wireless data transmission device. The software was developed to filter and process the data, derive the measures, and assess the risk of falling. Compared with available systems in the market, our inertial sensor based prototype system was very promising in terms of powerful functions, portability and low-cost on assessing fall risk of the older people.

The findings from this study and the developed prototype system could be incorporated into clinical practice to reliably identify “at-risk” individuals and to diagnose the underlying risk factors of falls in advance so that appropriate interventions can be implemented to reduce elderly people’s risk of falling. Such a system could improve their quality of life and reduce costs in the healthcare system.