Wearable Gait Biomarkers and Explainable AI Identify High Prodromal Burden in Parkinson's Disease

Introduction

This database includes the raw data associated with the manuscript “Wearable Gait Biomarkers and Explainable AI Identify High Prodromal Burden in Parkinson’s Disease”.

The dataset was designed to support the identification of biomechanical gait signatures associated with different levels of prodromal burden in Parkinson’s disease. It focuses on trunk kinematic features extracted from wearable inertial sensors during walking tasks.

Methods

The dataset includes anonymized gait features extracted from trunk-mounted inertial measurement units (IMUs) recorded during standardized walking assessments.

Each row corresponds to a single observation and is identified by a unique, non-informative record identifier. No personal or re-identifiable information is included.

Extracted features include spatio-temporal parameters, trunk acceleration-derived metrics, and entropy-based measures reflecting gait variability and complexity. Clinical labels related to prodromal burden classification are also provided.

Results

The dataset enables the analysis of gait-related biomechanical patterns associated with increasing prodromal burden. It supports machine learning–based classification and explainable AI approaches aimed at identifying clinically meaningful trunk gait signatures.