Sage Bionetworks - Gait Feature Extraction using PDKit & Rotation Features

Author: Aryton Tediarjo, Larsson Omberg

About

This repository is used for retrieving gait features from accelerometer and gyroscope data. It acts as a wrapper to an external Python package Parkinson Disease Featurization pacakge called pdkit with an additional QC handling of detecting rotation when user is walking based on rotation detection paper.

Added Functionality

• Additional Error Handling
• Added windowing feature computation
• Added QC on detecting rotation

Installation

Requirements:

Python 3 and above

Installing as a Github Python Package

pip install git+https://github.com/arytontediarjo/PDKitRotationFeatures.git

Use Case

from PDKitRotationFeatures import gait_module  
featureObjs = gait_module.GaitFeatures(window_size = 256, detect_rotation=True) ##refer to module for additional parameter
featureObjs.run_pipeline(accel sensor data <t,x,y,z>, gyro sensor data <t,x,y,z>)

Feature Dictionary

In the feature processing steps, the module will try to detect rotation given a gyroscope information. After separating the rotation sequence from the longitudinal data, features will be computed in smaller windows (default to 512 ~ around 5 seconds given 100Hz sampling frequency).

These are the features that will be computed:

1. Number of Steps: The number of steps in a given window

2. Cadence: The number of steps per window duration

3. Energy Freeze Index: Freeze Index is defined as the power in the “freeze” band [3–8 Hz] divided by the power in the “locomotor” band [0.5–3 Hz] [measured in Hz]

4. Locomotor Freeze Index: Locomotor freeze index is the power in the “freeze” band [3–8 Hz] added to power in the “locomotor” band [0.5–3 Hz]. [measured in Hz]

5. Stride-to-Stride/Step-to-Step Features: Compute step-to-step or stride-to-stride summary statistics (standard deviation, average etc.)

6. Step/Stride Regularity: Measure of step/stride regularity along axis [percentage consistency of the step-to-step pattern].

7. Speed of Gait: It extracts the gait speed from the energies of the approximation coefficients of wavelet functions. Preferably you should use the magnitude of x, y and z (mag_acc_sum) here, as the time series.

8. Gait Symmetry: Measure of gait symmetry along axis [difference between step and stride regularity].

9. Rotation Speed (Omega): Measures rotation speed based on inferred rotation period (radians during inferred rotation period/rotation duration)

References

1. Rotation-Detection Paper : https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5811655/
2. PDKit Docs: https://pdkit.readthedocs.io/en/latest/gait.html
3. PDKit Gait Source Codes : https://github.com/pdkit/pdkit