Data and algorithms used for the study: Exploring Unsupervised Feature Extraction of IMU-Based Gait Data in Stroke Rehabilitation using a Variational Autoencoder

Metadata Title/Name:

Dataset used for the paper: "Exploring Unsupervised Feature Extraction of IMU-Based Gait Data in Stroke Rehabilitation using a Variational Autoencoder"

Description:

Dataset contains inertial measurement unit based 2-minute walk test data of healthy individuals and individuals after stroke. The data consists both of test-retest data as well as longitudinal data. Per measurement, three sensors were used. Two sensors were placed on the feet, and one sensor on the lower back. The sensors measured with a frequency of 104 samples/s. The sensors contained an accelerometer and gyroscope.

Source:

Data was collected for the research project: Making sense of sensordata for personalised healthcare. The data was collected in 5 different rehabilitation centers.

Type:

The data consists of IMU measurements of gait measured during a 2-minute walk test.

Format:

The files are in CSV-format.

Structure:

The data was structured as follows: Subject_Status_TestType_Measurement_WalkingAid_Location.csv Subject = Subject number Status = Healthy/stroke TestType = Test_Retest / Longitudinal Measurement = Test/Retest/Tn (T = timepoint, n = 0,1,2,..) Walkingaid = Use of walking aid during assessment Location = Sensor placement

Content:

Sensordata contains a gait measurement.

Use Cases:

See Paper description below.

Test-retest data can be used to compute reliability. Data healthy vs stroke can be used to explore differences. Data longitudinal can be used to measure differences over time. Access and Usage:

Data can be used with the MIT license.

Relevant Links:

Felius, R. A. W., Geerars, M., Bruijn, S. M., van Dieën, J. H., Wouda, N. C., & Punt, M. (2022). Reliability of IMU-Based Gait Assessment in Clinical Stroke Rehabilitation. Sensors (Basel, Switzerland), 22(3), 908. https://doi.org/10.3390/s22030908

#Introduction

To gain a more comprehensive understanding of gait recovery, monitor progress and tailor interventions, measuring the way people walk is crucial [1], [2], [3]. Inertial Measurement Units (IMUs) are small and portable sensors that enable objective and continuous measurements of the way people walk. However, IMU data needs to be processed to extract relevant information before it can be used in research and clinical practice.

This study explored a data-driven approach of processing IMU data using Variational AutoEncoder (VAE) [4]. A VAE is a generative model that employs deep learning techniques to learn a compact, low-dimensional representation of data.

Variational AutoEncoder The VAE comprises two main components: an encoder and a decoder. The encoder maps the input-data to a lower-dimensional representation, known as the latent layer, by encoding it into a mean and variance vector. This vector is then used to generate a sample from a probability distribution that models the latent layer. The decoder takes this sample as input and generates a reconstructed output that is similar to the original input data. The difference between the original input and the reconstructed output is measured using a loss function. The VAE aims to minimize this loss function, which encourages the encoder to learn a good representation of the input data in the latent layer.

The input and output of the VAE consisted of a 512X6 epoch. The encoder and decoder both comprised three convolutional layers. The latent layer contained 12 latent variables.

References Sung Shin, Robert Lee, Patrick Spicer, and James Sulzer. Does kinematic gait quality improve with functional gait recovery? a longitudinal pilot study on early post-stroke individuals. Journal of Biomechanics, 105:109761, 03 2020. Elizabeth Wonsetler and Mark Bowden. A systematic review of mechanisms of gait speed change post-stroke. part 1: spatiotemporal parameters and asymmetry ratios. Topics in Stroke Rehabilitation, 24:1–12, 02 2017. Michiel Punt, Sjoerd Bruijn, Kim van Schooten, Mirjam Pijnappels, Ingrid Port, Harriet Wittink, and Jaap Van Dieen. Characteristics of daily life gait in fall and non fall-prone stroke survivors and controls. Journal of NeuroEngineering and Rehabilitation, 13, 07 2016. Diederik Kingma and Max Welling. An Introduction to Variational Autoencoders. 01 2019. (2024-04-16)