# The Use of a Smartphone to Assess the Two-Minute Step Test: Validity of Machine Learning Compared to Analytical Data Processing

## TL;DR

A smartphone held to the thigh can accurately measure movement during a 2-minute step test, with machine learning providing better results than traditional methods.

## Contribution

The study shows that machine learning improves smartphone-based assessment of thigh movement during the 2MST compared to analytical methods.

## Key findings

- Smartphone-held thigh measurements matched motion capture for step count and timing.
- Machine learning provided more accurate peak thigh angular velocity estimates than analytical methods.
- Analytical methods had around 8% error compared to ground truth motion capture data.

## Abstract

What are the main findings?

A smartphone held to the thigh provided valid estimates of several parameters of the 2MST.

A machine learning data processing approach provided peak thigh angular velocity with better agreement to the ground truth measure than an analytical data processing approach.

What are the implications of the main findings?

Holding a smartphone to the thigh provides a simple, easy, and low-cost way to measure 2MST parameters.

The better agreement for machine learning-determined thigh peak angular velocity suggests that such an approach should be considered when using a smartphone to determine 2MST parameters.

The 2-Minute Step Test (2MST) is commonly scored by step count, which overlooks how the task is performed. This study tested whether a smartphone held to the thigh can be used to quantify thigh kinematics to determine 2MST outcome parameters, and whether a machine learning (ML) data analysis approach of the smartphone signal yields better agreement with motion capture (ground truth) compared to a more typical analytical data analysis approach (AA). Eighty-four healthy adults completed the 2MST while holding a smartphone against the right thigh. A thigh angular velocity ‘ground truth’ reference was obtained by simultaneous recording via motion capture (Vicon). Smartphone signals were resampled and processed using analytical (i.e., adaptive Butterworth filtering) and machine-learning data processing approaches (i.e., a stacked regression model trained to identify peak angular velocities). Step cycles and cycle duration were identical across equipment modalities and data analysis pipelines (mean 143 ± 18 cycles; 0.84 ± 0.11 s). However, the mean and variability of peak thigh angular velocity differed across the different modalities/pipelines (motion capture: 303 ± 39°·s−1; AA: 280 ± 47°·s−1; ML: 304 ± 37°·s−1). Bland–Altman agreement, compared to the ground truth measure, showed larger bias and limits of agreement for AA (bias 25.5°·s−1; −49.8–100.8) compared to ML (bias 1.0°·s−1; −15.4–17.5). These findings support using a smartphone held to the thigh to assess how the 2MST is performed, including providing the number and timing of steps completed and the average and variability in thigh angular velocity across cycles. Findings also suggest that a machine learning data analysis approach provides thigh angular velocity measures that are nearly identical to motion capture techniques, whereas a typical analytical data analysis approach has errors of around 8%.