Wearable Gait Signals from Parkinson and Healthy Subjects Using Foot IMUs

General overview

This repository contains the data collected during the MyGait project, which aims to improve the detection and monitoring of gait abnormalities in individuals with Parkinson’s disease through the use of wearable inertial sensors and machine learning algorithms.

Inertial signals are collected from sensors placed on the participants’ footwear while they perform various standardized physical tests. These data enable the training and validation of models capable of identifying pathological gait patterns.

Data Structure

The total size of the dataset amounts to approximately 2.82 GB, consisting of CSV files that record data at a frequency of 50 Hz. The recordings come from six-axis inertial sensors (IMUs), including triaxial accelerometers and gyroscopes, mounted on the outer side of the footwear.

Each line of the CSV files contains 21 numerical columns (float32):

• timestamp: timestamp of each sample.
• sensor_id: sensor identifier.
• foot: corresponding foot (left or right).
• seq: sequence number.
• acx, acy, acz: acceleration along the three axes.
• gyrx, gyry, gyrz: angular velocity along the three axes.
• roll, pitch, yaw: angular orientation.
• p_...: plantar pressure data (not used due to reliability issues).

The data are organized into folders by participant group (individuals with Parkinson’s / healthy controls) and by the type of test performed. The folder names reflect both the order in which the tests were conducted and the specific type of test.

Files names

Each CSV file follows the naming format:

[ID]-RTSOCS - YYYY.MM.DD-HH.MM.SS.csv

Where:

• ID refers to the participant’s identifier within the study. For tests conducted outdoors, each repetition is distinguished by the suffix -1 or -2.
• YYYY.MM.DD indicates the date the data were recorded.
• HH.MM.SS indicates the start time of the recording.

This format allows for precise identification of when each recording was performed, as well as which subject and repetition it corresponds to. The files are organized within folders according to the type of test.

In the 6min_out test (six-minute walk outdoors), each participant performed the test twice. This repetition is indicated in the filename by the suffix -1 or -2 in the participant identifier, making it easy to distinguish between the two sessions.

Participants and Data Collection

The study included 44 individuals with Parkinson’s disease and 45 healthy controls, recruited through the Parkinson Aragón Association and public outreach. The tests were conducted between October 2022 and December 2023 in both indoor settings (a marked hallway) and outdoor environments (a courtyard with uneven terrain and a ramp), under the supervision of clinical staff.

Each participant performed the following tests:

• 10 Meter Walk Testing Form
  • 10mLento (10mSlow): walking at a comfortable pace
  • 10mRapido (10mFast): walking at maximum speed
• Mini-BESTest (balance assessment):
  • Anticipatory: ANTStandUp, ANTTiptoes, ANTSingleLegRight, ANTSingleLegLeft.
  • Reactive Postural Control: REACTStepForward, REACTStepBackward, REACTStepRight, REACTStepLeft.
  • Sensory Orientation: SENSOpenEyes, SENSSoftFoam, SENSSlope.
  • Dynamic Gait: DGChangeSpeed, DGHeadTurns, DGPivotTurn, DGObstacle, DGTUG (Timed Up and Go), DGTUGTask (TUG with cognitive or dual task).
• 6-Minute Walk Test
  • 6min_in: test performed indoors
  • 6min_out: test performed outdoors

During these tests, clinical observations (such as fatigue, dyspnea, and number of laps) were recorded and they are included in an excel file.

Folder Structure

Parkinson Patients:

• 1_10mSlow
• 2_10mFast
• 3_DGChangeSpeed
• 4_DGHeadTurns
• 5_DGPivotTurn
• 6_DGObstacle
• 7_DGTUG
• 8_DGTUGTask
• 9_ANTStandUp
• 10_ANTTiptoes
• 11_ANTSingleLegRight
• 12_ANTSingleLegLeft
• 13_REACTStepForward
• 14_REACTStepBackward
• 15_REACTStepRight
• 16_REACTStepLeft
• 17_SENSOpenEyes
• 18_SENSSoftFoam
• 19_SENSSlope
• 20_6min_in
• 21_6min_out

Healthy patients:

• 1_10mSlow
• 2_10mFast
• 3_DGChangeSpeed
• 4_DGHeadTurns
• 5_DGPivotTurn
• 6_DGObstacle
• 7_DGTUG
• 8_DGTUGTask
• 9_6min_in
• 10_6min_out

Each folder contains CSV files with the recordings corresponding to that specific test, differentiated by foot (left or right) and participant code.

A complementary Excel file is also included, containing anonymized sociodemographic and clinical information for the participants used in the main analysis (such as age, sex, group, etc.). This information is referenced using the same participant codes that appear in the CSV filenames.