# A multi-level annotated sensor dataset of gait freezing manifestations and severity in Parkinson’s disease

## TL;DR

This paper introduces FoG-STAR, a sensor dataset for studying gait freezing in Parkinson’s disease, annotated by experts for algorithm development.

## Contribution

The paper introduces a multi-level annotated sensor dataset for studying freezing of gait in Parkinson’s disease.

## Key findings

- The dataset includes 101 annotated freezing of gait episodes from 22 participants performing seven motor tasks.
- Tri-axial IMU data and expert annotations enable analysis of FoG in dynamic motor contexts.
- The dataset supports machine learning for FoG detection, severity assessment, and activity recognition.

## Abstract

We present FoG-STAR, a dataset collected using wearable sensors, designed to support the development and evaluation of algorithms for detecting and characterizing freezing of gait (FoG) in people with Parkinson’s disease (PD). The dataset includes recordings from 22 participants who performed a series of standardized motor tasks while wearing four inertial measurement units (IMUs) on the ankles, wrist, and lower back. Each IMU recorded tri-axial accelerometer and gyroscope data. Participants completed seven structured tasks, including walking with/without cognitive/motor dual-tasks, 360-degree turning, and the timed-up-and-go test, which comprises six types of activities (sitting, standing, sit-to-stand, stand-to-sit, walking, and turning). The dataset features detailed annotations from two expert clinical raters, who marked the onset and offset of 101 FoG episodes, and labelled specific FoG manifestations. In addition, the duration of each activity and task segment was annotated. This multi-level annotation framework allows for studying FoG in the context of dynamic motor behavior and provides a valuable resource for the development of machine learning models aimed at FoG detection, severity assessment, and activity recognition in PD.