TSDF: A simple yet comprehensive, unified data storage and exchange format standard for digital biosensor data in health applications

TL;DR

This paper introduces TSDF, a unified data format for physiological biosensor data, designed to improve data storage, exchange, and reproducibility across health applications by standardizing data and metadata formats.

Contribution

The paper proposes a simple, comprehensive standard for biosensor data storage and exchange, addressing gaps in current formats with a focus on usability and interpretability.

Findings

Current formats lack comprehensiveness and standardization.

TSDF uses raw binary and JSON for data and metadata.

Standardization enhances data interpretability and reproducibility.

Abstract

Digital sensors are increasingly being used to monitor the change over time of physiological processes in biological health and disease, often using wearable devices. This generates very large amounts of digital sensor data, for which, a consensus on a common storage, exchange and archival data format standard, has yet to be reached. To address this gap, we propose Time Series Data Format (TSDF): a unified, standardized format for storing all types of physiological sensor data, across diverse disease areas. We pose a series of format design criteria and review in detail current storage and exchange formats. When judged against these criteria, we find these current formats lacking, and propose a very simple, intuitive standard for both numerical sensor data and metadata, based on raw binary data and JSON-format text files, for sensor measurements/timestamps and metadata, respectively. By focusing on the common characteristics of diverse biosensor data, we define a set of necessary and sufficient metadata fields for storing, processing, exchanging, archiving and reliably interpreting, multi-channel biological time series data. Our aim is for this standardized format to increase the interpretability and exchangeability of data, thereby contributing to scientific reproducibility in studies where digital biosensor data forms a key evidence base.