欧美亚洲国产精品久久高清,国产精品欧美久久久久无广告,99精品欧美一区二区三区

                                                  跳到正文

                                                  Hitachi

                                                  研究開發

                                                  Industrial AI Blog from China

                                                  Applying motor ability tests to fall risk assessment to help improve QoL for the elderly

                                                  9 September 2021

                                                  Ma Jian

                                                  Jian Ma
                                                  Hitachi (China) Research & Development Corporation

                                                  Why we need to look at fall risk

                                                  According to WHO’s study [1], falls are one of the top causes of diseases, which leads to 684 deaths per million population worldwide. It affects elderly’s quality-of-life (QoL) most due to their degenerated physical and mental conditions and puts heavy burden on their families and society at large. In China, the aging population is also a priority issue of the society. UN estimates that there will be over 480 million Chinese people over the age of 60 by 2050 [2]. According to Chinese Center for Disease Control and Prevention (CDC) fall injury is the top cause of deaths to the elderly over 65 and about 40 million elderly people fall at least once annually [3]. How to deliver better elderly care and manage this disease is a big challenge, and is key to successful elderly care systems. Developing automated fall risk assessment is a good solution to this problem and will provide a powerful instrument to healthcare providers.

                                                  Previously, we developed two technologies for smart elderly care: one for predicting dementia [4] and the other for automatic fall risk assessment [5]. The former predicts Minimal Mental State Examination (MMSE) score -- a most popular clinical tool for dementia screening -- from a group of selected characteristics of finger tapping movement with predictive models. The latter predicts Timed Up and Go (TUG) score -- a widely used fall risk assessment tool -- from gait characteristics derived from video with deep learning and stereo vision. These two studies were based on two types of data collected from a finger-tapping test and TUG test, respectively, and these two types of data were analyzed independently in previous research. However, as one might expect, cognitive and functional disorders usually occur together in aging people, and there is scientific evidence indicating that cognition impairment can help to predict fall risk [6] and that cognition and gait are associated with each other [7]. In previous studies, we discovered the relationship between certain characteristics of finger tapping and cognition impairment measured by MMSE score [4]. This time, we decided to see if a relationship exists between finger tapping, cognition, gait and fall risk if we jointly analyzed all the data together.

                                                  The mathematical tool that we used: Copula Entropy

                                                  We used a mathematical tool called Copula Entropy (CE) defined by Ma and Sun [8]. It is an ideal mathematical concept for statistical independence testing and enjoys several axiomatic properties for independence measure which other independence measures don’t have. Compared with traditional bi-variate statistical correlation measures, such as Pearson Correlation Coefficient, which is only for Gaussian cases, it can be applied to measure any kinds of multivariate correlations (linear/nonlinear) without any assumptions on the underlying distributions. A simple method for estimating CE is also available which makes our analysis easy and sound. It should be mentioned that CE is also the tool we used in the previous studies on the relationship between finger motor and cognitive ability [4] and on the relationship between gait characteristics and fall risk [5].

                                                  How and what data we collected

                                                  The data used in this research was collected from the people who performed MMSE test, TUG test, and finger-tapping test for one month at China. For each person, the characteristics of finger-tapping movement were derived from finger-tapping test and gait characteristics from TUG test. MMSE score and TUG score are collected at the same time.

                                                  The experiments we did on the collected data

                                                  To study the relationship between finger-tapping, gait, and fall risk, we conducted an experiment to measure the associations between scores, the characteristics of finger tapping and gait with CE. We then used the discovered relationships to improve our technology for automatic fall risk assessment.

                                                  Figure1
                                                  Figure 1. Associations of TUG score with the characteristics of finger tapping (red) and gait (blue)

                                                  The associations between TUG score and all the characteristics are shown in Figure 1. We can learn that 9 characteristics of finger-tapping (‘number of taps’, ‘average interval of taps’, ‘frequency of tapping’ of both hands of bimanual in-phase movement (lnumber1, rnumber1, lavgint1, ravgint1, lfreq1, rfreq1), and of left hands of bimanual unti-phase movement (lnumber2, lavgint2, lfreq2) are associated with TUG score with relatively high strength compared with gait characteristics.

                                                  These interesting results suggest that finger motor ability and functional ability may be related in elderly people. To the best of our knowledge, this is the first research reporting such a relationship. It also inspired us to use these characteristics of finger-tapping to improve the models that predict TUG score with only gait characteristics.

                                                  To check whether such improvement is possible, we conducted comparison experiments by building predictive models with three groups of characteristics (6 characteristics of both hands of bimanual in-phase movement, gait only, and both) as inputs. The performance of two models in terms of MAE are listed in Table 1. The predictive models in the experiments are Linear Regression (LR) and Support Vector Regression (SVR) – two of the most widely used machine learning models. We found that gait characteristics can present stable performance for both models, and that SVR with both characteristics as inputs presents the best results as shown in Figure 2 (MAE = 1.160). This means that adding the associated characteristics of finger-tapping into the inputs of models can make more accurate predictive models for fall risk assessment than before.

                                                  Table 1. MAE of two models with three groups of inputs

                                                  Finger tapping Gait Both
                                                  LR 44.586 1.341 15.050
                                                  SVR 1.365 1.201 1.160

                                                  As a summary, from the joint data set of scores and characteristics, we discovered that a relationship exists between finger motor ability and functional ability in aging people. We then utilized such associations to build a better model for automatic fall risk assessment by adding the associated characteristics of finger-tapping into the inputs of the model.

                                                  Discovering such relationships is of great value both from the perspective of science as well as technology because it will not only deepen our biomedical understanding of aging but also enable us to develop better technologies for elderly care as the associations have biomedical meaning, and the model built on such associations is meaningful to clinical users, and thereby useful in real world applications.

                                                  For more details, please refer to our preprint paper “Associations between finger tapping, gait and fall risk with application to fall risk assessment ” on arXiv [9].


                                                  Figure 2. Predictive performance of the SVR models with 3 groups of characteristics


                                                  References

                                                  [1]
                                                  World Health Organization. Global Health Estimates Summary Tables: Deaths by Cause, Age and Sex by various regional grouping 2000-2019. December 2020.
                                                  [2]
                                                  United Nations (UN), World Population Prospects 2019, https://population.un.org/wpp/DataQuery/.
                                                  [3]
                                                  Chinese Center for Disease Control and Prevention. Notice on elderly’s fall prevention. June 11st, 2019.
                                                  [4]
                                                  Jian Ma. Predicting MMSE Score from Finger-Tapping Measurement. arXiv preprint, arXiv:2004.08730, 2020.
                                                  [5]
                                                  Jian Ma. Predicting TUG score from gait characteristics with video analysis and machine learning. arXiv preprint, arXiv:2003.00875, 2020.
                                                  [6]
                                                  Susan W Muir, Karen Gopaul, and Manuel M Montero Odasso. The role of cognitive impairment in fall risk among older adults: a systematic review and meta-analysis. Age and ageing, 2012, 41(3): 299-308.
                                                  [7]
                                                  Nancye May Peel, Linson John Alapatt, Lee Vanessa Jones, Ruth Eleanor Hubbard. The association between gait speed and cognitive status in community-dwelling older people: a systematic review and meta-analysis. The Journals of Gerontology: Series A, 2019, 74(6): 943-948.
                                                  [8]
                                                  Jian Ma and Zengqi Sun, Mutual information is copula entropy. Tsinghua Science & Technology, 2011, 16(1): 51-54. See also arXiv preprint arXiv:0808.0845, 2008.
                                                  [9]
                                                  Jian Ma. Associations between finger tapping, gait and fall risk with application to fall risk assessment. arXiv preprint arXiv:2006.16648, 2020.
                                                  欧美亚洲国产精品久久高清,国产精品欧美久久久久无广告,99精品欧美一区二区三区