Quantitative falls risk estimation through multi-sensor assessment of standing balance

Abstract

Falls are the most common cause of injury and hospitalization and one of the principal causes of death and disability in older adults worldwide. Measures of postural stability have been associated with the incidence of falls in older adults. The aim of this study was to develop a model that accurately classifies fallers and non-fallers using novel multi-sensor quantitative balance metrics that can be easily deployed into a home or clinic setting. We compared the classification accuracy of our model with an established method for falls risk assessment, the Berg balance scale. Data were acquired using two sensor modalities--a pressure sensitive platform sensor and a body-worn inertial sensor, mounted on the lower back--from 120 community dwelling older adults (65 with a history of falls, 55 without, mean age 73.7 ± 5.8 years, 63 female) while performing a number of standing balance tasks in a geriatric research clinic. Results obtained using a support vector machine yielded a mean classification accuracy of 71.52% (95% CI: 68.82-74.28) in classifying falls history, obtained using one model classifying all data points. Considering male and female participant data separately yielded classification accuracies of 72.80% (95% CI: 68.85-77.17) and 73.33% (95% CI: 69.88-76.81) respectively, leading to a mean classification accuracy of 73.07% in identifying participants with a history of falls. Results compare favourably to those obtained using the Berg balance scale (mean classification accuracy: 59.42% (95% CI: 56.96-61.88)). Results from the present study could lead to a robust method for assessing falls risk in both supervised and unsupervised environments.

Review of fall risk assessment in geriatric populations using inertial sensors

“…Of the variables that were assessed in more than one study, only 13 were significant (p < 0.05) each time they were assessed: 1) mediolateral and anteroposterior postural sway length [43,51]; 2) mediolateral and anteroposterior postural sway velocity [43,51]; 3) ratio of mean squared modulus for postural sway [46-48]; 4) standard deviation of anteroposterior acceleration [18,74]; 5) root mean square amplitude of vertical linear acceleration [55,72]; 6) gait speed [40,44,55,64,72,74]; 7) sit-to-stand transition duration [17,45]; 8) dominant Fast Fourier Transform (FFT) peak parameters derived from lower-back linear acceleration signals [59,61,62]; 9) ratio of even to odd harmonic magnitudes derived from head, upper back, and lower-back linear acceleration signals [44,55,57,58,64,72,75]; 10) area under the first six harmonics divided by the remaining area for lower-back linear acceleration signals [57,58]; 11) ratio of the first four harmonics to the magnitude of the first six harmonics for lower-back linear acceleration signals [57,58]; 12) maximum Lyapunov exponent of angular velocity signal [73,77]; 13) discrete wavelet transform parameters from lower-back angular velocity and linear acceleration signals and sternum linear acceleration signals [45,63]. Six of these multi-study variables (1,2,5,6,9,12) were from different research groups, while seven variables (3,4,7,8,10, 11,13) were from a single research group.…”

“…Mediolateral and anteroposterior postural sway length and velocity are measures of postural stability that represent trunk movement during static standing [43,51]. The root mean square of vertical linear accelerations has been used to measure gait smoothness, with larger values linked to increased fall risk [55,72].…”

“…The five best performing models, in terms of overall accuracy, specificity, and sensitivity, used neural networks [47,48], naive Bayesian classifier [41], Mahalanobis cluster analysis [46], and a decision tree [49]. The five worst performing models used regression [44,53,62,72] and a support vector machine [51]. Therefore, intelligent computing methods (neural networks, Bayesian classifiers, etc.)…”

“…Several papers emphasized the importance of using prospective fall risk occurrence in their future research [51,54,60,62,72-74], thereby avoiding the two biggest limitations of retrospective fall assessment: inaccurate recollection of fall history and gait changes due to past falls.…”

Using a Motion Sensor to Categorize Low Back Pain Patients: A Machine Learning Approach

“…Since there was no similar study designed to classify LBP patients such that there are different treatment approaches for each class, we then we compared our model to other machine-learning-based models that were developed for clinical decision making [45][46][47]. The obtained accuracy for the best machine learning model (i.e., the SVM algorithms) in this study was in the range of clinical decision-making models described in the literature that were based on signals from IMU or body-worn sensors (i.e., <80%) [45][46][47][48]. Compared to the SVM model, the MLP approach achieved a lower accuracy of ~60%, which could be related to the fairly small sample size in this study (94 for 3-class and about 65 for 2-class).…”