Criterion Validity of Linear Accelerations Measured with Low-Sampling-Frequency Accelerometers during Overground Walking in Elderly Patients with Knee Osteoarthritis

Abstract

Sensors with a higher sampling rate produce higher-quality data. However, for more extended periods of data acquisition, as in the continuous monitoring of patients, the handling of the generated big data becomes increasingly complicated. This study aimed to determine the validity and reliability of low-sampling-frequency accelerometer (SENS) measurements in patients with knee osteoarthritis. Data were collected simultaneously using SENS and a previously validated sensor (Xsens) during two repetitions of overground walking. The processed acceleration signals were compared with respect to different coordinate axes to determine the test-retest reliability and the agreement between the two systems in the time and frequency domains. In total, 44 participants were included. With respect to different axes, the interclass correlation coefficient for the repeatability of SENS measurements was [0.93-0.96]. The concordance correlation coefficients for the two systems' agreement were [0.81-0.91] in the time domain and [0.43-0.99] in the frequency domain. The absolute biases estimated by the Bland-Altman method were [0.0005-0.008] in the time domain and [0-0.008] in the frequency domain. Low-sampling-frequency accelerometers can provide relatively valid data for measuring the gait accelerations in patients with knee osteoarthritis and can be used in the future for remote patient monitoring.

Keywords: SENS sensors; frequency-domain comparison; gait accelerations; inertial measurement units; knee osteoarthritis; low-sampling-frequency accelerometers; remote monitoring of patients; test–retest reliability; time-domain comparison; wearable motion-tracking sensors.

Conflict of interest statement

The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

Figures

Figure 1

(A) Simultaneous application of SENS and Xsens sensors (Xsens sensors are shown on top of clothing for illustrative purposes). (B) Placement of a SENS sensor on the distal lateral side of the thigh. (C) Coordinate axes of the SENS sensor.

Figure 2

Reconstruction of continuous SENS signal using the Fourier method. The gray dots show the accelerations recorded by the SENS sensors along the x-axis as an example of one gait cycle. The blue line shows the continuous signal reconstructed using the Fourier method.

Figure 3

An overview of the signal processing protocol used in this study: (A) Simultaneous recording of the gait signals obtained with SENS and Xsens during two overground gait trials. (B) Inspection of the PSD (power spectral density) of the signals and filtering of the signals after determining a cutoff frequency of 4 Hz using a fourth-order zero-lag low-pass Butterworth filter. (C) Temporal matching of the signals using the cross-correlation method. (D) Segmentation of the gait into five individual gait cycles. (E) Averaging and normalization of the gait cycles into gait cycle percentages with respect to different coordinate axes and the magnitude vector.

Figure 4

(Left column): scatterplots demonstrating the correlation between the SENS and Criterion accelerations with respect to different axes. The black line depicts the line of equality (SENS acc = Criterion acc). (Right column): Bland–Altman plots demonstrating the agreement between the SENS and Criterion accelerations. The upper and lower LoAs (limits of agreement) are shown as the upper and lower dashed lines corresponding to the 2.5th and 97.5th percentiles of the differences. (Data from the right-side sensor are marked in blue, and those from the left-side sensor is marked in red).

Figure 5

(Left column): scatterplots demonstrating the correlation between the frequencies of the peaks of the PSDs of SENS and Criterion with respect to different axes. The solid black line depicts the line of equality (SENSFreq == CriterionFreq). (Right column): corresponding Bland–Altman plots demonstrating the agreement between the frequencies of the peaks of the PSDs of SENS and Criterion. The upper and lower LoAs are shown as the upper and lower dashed lines corresponding to the 2.5th and 97.5th percentiles of the differences. (Data from the right-side sensor are marked in blue, and those from the left-side sensor are marked in red).

Figure 6

(Left column): scatterplots demonstrating the correlation between the powers of the peaks of the PSDs of SENS and Criterion with respect to different axes. The solid black line depicts the line of equality (SENSpower == Criterionpower). (Right column): corresponding Bland–Altman plots demonstrating the agreement between the powers of the peaks of the PSDs of SENS and Criterion. The upper and lower LoAs are shown as the upper and lower dashed lines corresponding to the 2.5th and 97.5th percentiles of the differences. (Data from the right-side sensor are marked in blue, and those from the left-side sensor are marked in red).

Figure 7

(Left column): scatterplots demonstrating the correlation between the Fourier coefficients of SENS and Criterion with respect to different axes. The solid black line depicts the line of equality (SENS coefficients == Criterion coefficients). (Right column): corresponding Bland–Altman plots demonstrating the agreement between the Fourier coefficients of the peaks of the PSDs of SENS and Criterion. The upper and lower LoAs are shown as the upper and lower dashed lines corresponding to the 2.5th and 97.5th percentiles of the differences. (Data from the right-side sensor are marked in blue, and those from the left-side sensor are marked in red).