Clinical Dataset Structure: A Universal Standard for Structuring Clinical Research Data and Metadata

Abstract

Purpose : During clinical research studies, multiple modalities of data are typically collected such as surveys, vitals, and eye images. There is currently no consensus on how to structure such multimodal data into a consistently organized dataset that is easily reusable by humans and machines in line with the FAIR (Findable, Accessible, Interoperable, Reusable) Principles. We addressed this issue in the Artificial Intelligence Ready and Equitable Atlas for Diabetes Insights (AI-READI) project by developing the Clinical Dataset Structure (CDS), a standard approach for organizing multimodal clinical research data and metadata at the root level.

Methods : We first reviewed modality-specific standard structures such as the Brain Imaging Data Structure (BIDS) for neuroimaging data and the SPARC Data Structure (SDS) for neuromodulation-related data to learn from their approach. We simultaneously reviewed popular metadata schemas such as the DataCite and ClinicalTrials.gov schemas to identify critical metadata elements and ways to structure them. We also reviewed artificial intelligence (AI) and machine learning (ML) specific metadata standards such as datasheet for dataset and data card. Using all that information, we established the first version of the CDS. We applied the CDS to the pilot data from the AI-READI study and evaluated it for alignment with FAIR Principles relevant to data and metadata structures.

Results : The CDS mandates to organize data into one folder per datatype, named as per a set convention, where applicable standard structure must then be followed within each folder. The CDS requires several human-friendly metadata files to be included at the root-level such as a README file that provides an overview, a standard datasheet, and a LICENSE file with reuse terms. Machine-friendly metadata files are specified as well such as a DataCite-compliant dataset description file and a ClinicalTrials.gov-inspired study description file. Evaluation of the AI-READI pilot dataset showed that the CDS makes datasets compliant with all the relevant elements from the FAIR Principles.

Conclusions : The CDS provides a simple and intuitive way to organize clinical research data and metadata at the root level such as to optimize their interoperability and reusability, especially for AI/ML applications.

This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.

 

Illustration of a sample dataset organized at the root level according to the CDS.