Random forest algorithms to classify frailty and falling history in seniors using plantar pressure measurement insoles: a large-scale feasibility study

Abstract

Background: Frailty and falls are two adverse characteristics of aging that impair the quality of life of senior people and increase the burden on the healthcare system. Various methods exist to evaluate frailty, but none of them are considered the gold standard. Technological methods have also been proposed to assess the risk of falling in seniors. This study aims to propose an objective method for complementing existing methods used to identify the frail state and risk of falling in older adults.

Method: A total of 712 subjects (age: 71.3 ± 8.2 years, including 505 women and 207 men) were recruited from two Japanese cities. Two hundred and three people were classified as frail according to the Kihon Checklist. One hundred and forty-two people presented with a history of falling during the previous 12 months. The subjects performed a 45 s standing balance test and a 20 m round walking trial. The plantar pressure data were collected using a 7-sensor insole. One hundred and eighty-four data features were extracted. Automatic learning random forest algorithms were used to build the frailty and faller classifiers. The discrimination capabilities of the features in the classification models were explored.

Results: The overall balanced accuracy for the recognition of frail subjects was 0.75 ± 0.04 (F1-score: 0.77 ± 0.03). One sub-analysis using data collected for men aged > 65 years only revealed accuracies as high as 0.78 ± 0.07 (F1-score: 0.79 ± 0.05). The overall balanced accuracy for classifying subjects with a recent history of falling was 0.57 ± 0.05 (F1-score: 0.62 ± 0.04). The classification of subjects relative to their frailty state primarily relied on features extracted from the plantar pressure series collected during the walking test.

Conclusion: In the future, plantar pressures measured with smart insoles inserted in the shoes of senior people may be used to evaluate aspects of frailty related to the physical dimension (e.g., gait and balance alterations), thus allowing assisting clinicians in the early identification of frail individuals.

Keywords: Aging; Balance; Fall risk; Frailty; Gait analysis; Plantar pressure; Random forest classifier; Smart-insole; Walking.

Fig. 1

Overview of the 7-sensor plantar pressure measurement insole device and output. A: Seven pressure sensors are inserted in a 2 mm-thick hygienic shoe insole. The insole is inserted in a commercial Velcro shoe, and the data acquisition unit is attached using an additional piece of Velcro on the tong. B: A five-second example of plantar pressure data was collected from a walking test

Fig. 2

Overview of the data reduction process. AUC: area under the curve. COP: center of pressure

Fig. 3

Illustration of some selected parameters computed during the data feature extraction process. A: Data features extracted from the Standing COP analysis. The black line illustrates the COP excursion trajectory. The light blue area illustrates the surface covered by the COP excursion trajectory. B: Variables used in 1-foot COP trajectory analysis. The black line illustrates the COP trajectory. The red line segment illustrates the 1-foot COP excursion. The numbered black squares indicate the virtual locations of the 7 sensors. The green triangle marks are the starting point and the endpoint of the COP trajectory. C: Example of plantar pressure time series for one isolated step obtained during the walking test. Plantar pressures treated for each isolated step are the raw material for the extraction of all time domain features in the following categories: “peak analysis and area under the curves”, “1-foot COP trajectory analysis”, “gait phase analysis” and “Wavelet analysis”. Blue: heel, orange: lateral midfoot, green: center of the midfoot, red: lateral forefoot, purple: center of the forefoot, brown: medial forefoot, pink: big toe. D: Variables used in the wavelet analysis (extracted from C). The black line corresponds to the envelope of the 7 sensors. The blue triangle illustrates the first and the second peaks typically observed during the stance phase. The red diamond illustrates the valley between the two peaks. The orange lines describe the peak widths, calculated at 30% of their magnitude. The green break lines correspond to the slopes on each side of the peaks. x: medial-lateral axis. y: anterior-posterior axis. In panel A, COP excursion distances were doubled on the x and y axes to increase readability

Fig. 4

Performance of the frailty state and falling history classifiers for the whole population presented as confusion matrices. A: frailty state. B: falling history

Fig. 5

Performance of the frailty state classifiers presented as confusion matrices for all the subgroups. A: aged ≥65 years old. B: aged between 60 and 69 years old. C: aged between 70 and 74 years old. D: aged ≥75 years old. E: women only (≥ 65 years). F: men only (≥ 65 years)