Journal DOI 10.68956/journal.human-performance-futures

Creative Commons source reading · Data quality and algorithm evaluation

Transforming a clinical study database into a structured database adapted to artificial intelligence applications

Thibault Sauron; Carole Lazarus; Camille Kurtz; Florence Cloppet; Isabelle Thomassin Naggara; EURAD Study GroupPonceletEdouard5Jalaguier-CoudrayAurelie6MilletIngrid7JuhanValerie8BalleyguierCorinne9MalhaireCaroline10PerrotNicolas11BazotMarc12TaourelPatrice7DaraiEmile13; Laure Fournier

Originally published in Insights Imaging · 2026-02-16

From the article

Selected source text. JournalOS removes markup and shortens excerpts where indicated. The full article provides the complete context.

Abstract

ObjectiveMedical imaging databases suitable for training machine learning/computer vision algorithms are scarce, limiting the potential for development and generalisation of clinical tools. Clinical trial databases are a source of data, known for their high-quality data and reliable annotations. However, they are not tailored to the needs of machine learning or deep learning models. Our objective was to develop a methodology and tools that enable the curation of these databases specifically for the training or testing of artificial intelligence tools.Materials and methodsMRIs from the French centres of the EURAD clinical trial (MRI of women with pelvic adnexal lesions) were used to constitute the database. We developed the steps required to curate a clinical trial database: definition of inclusion and exclusion criteria, removal of…

Discussion

In this study, we established a process for transforming a clinical and imaging database from a clinical trial into an exploitable base for the development of AI tools and applied it to a specific clinical trial database. Our primary objective was to develop Content-Based Image Retrieval (CBIR); however, the database is designed to support the development of segmentation, classification, or data harmonisation tools as well. To do this, we had to develop curation and harmonisation tools to ensure regulatory and technical compliance. Grouped in a package called “DATACURATOR”, we provide these tools to the community in the form of open-source, freely available Python scripts. Our process was designed to be transposable to other clinical trial designs or imaging modalities, both in terms of the required steps and the critical issues to…

Conclusion

We developed freely accessible methods and tools embedded in an open-source package to facilitate the conversion of an existing medical database into a format suitable for training machine learning or computer vision algorithms. Our objective was to build upon the efforts deployed for a clinical trial and enable the development of AI tools based on the data and physician expertise. This work on MRI ovarian masses serves as a proof of concept, but a more systematic approach could be developed in the medical imaging community to offer guidelines and tools that would allow the successful transformation of clinical trial data into machine learning-compatible data on a large scale while remaining compliant with international regulations.

Editorial reading guide

  • What data created and tested the model, and was evaluation independent?
  • What is the cost of a false positive, false negative or out-of-distribution decision?