Sigmark: An Open-Source Tool for Collaborative Event Annotation in Wearable Sensor Data

Introduction
Wearable devices enable low-cost, real-world monitoring of individuals, capturing rich sensor data suitable for event detection. As studies increasingly rely on digital real-world monitoring, accurate and efficient event annotation (e.g., falls, disease-related symptoms) is essential. However, current labeling workflows often rely on manual logging or generic tools, which are time-consuming, error-prone, and lack the temporal precision required for high-frequency sensor data.

Methods
We developed Sigmark, an open-source graphical interface (Python, PySide6) designed to standardize the annotation of wearable sensor data (accelerometer, gyroscope, barometer). The application implements a rigorous two-stage workflow based on validated methodologies:

1. Independent Phase: Two or more reviewers independently annotate events and assign individual confidence scores using synchronized multi-sensor visualization.
2. Consensus Phase: Another reviewer imports all independent annotations to resolve conflicts and finalize verification certainty.

The event-related paper-based report is used as a reference to determine the visualization window (± 3 hours from the reported event time) and to display contextual information (e.g., fall description). Features include keyboard shortcuts for rapid navigation, signal-aggregation controls, visualization of task-related annotations (e.g., walking bouts, turns) derived from validated signal-based algorithms, and automated CSV export containing precise timestamps and metadata. 

Results
Sigmark was deployed to annotate fall events in the Mobilise-D clinical validation study, which used a single lower-back sensor. The tool successfully handled the import, annotation, and visualization of complex real-world data. Preliminary feedback indicates that the digital consensus workflow significantly improves annotation speed and temporal precision compared to traditional manual methods.

Conclusions
Sigmark provides an efficient, intuitive platform for annotating digital recordings of fall events captured by wearable sensors. Its general-purpose design also enables applications beyond falls, such as gait analysis. By releasing Sigmark as an open-source software, we aim to facilitate collaborative dataset creation and standardize annotation protocols across the research community.