Healthcare

Kereval is involved in a new research project called PénélopAlgo, which aims to facilitate the use of health data to improve patient targeting for clinical trials. The project addresses the challenge of low patient recruitment rates in clinical studies—which can hinder the validation of new therapies—as well as the limited access to clinical trials as a care option.

This project is led by AP-HP, in partnership with LIMICS (Laboratory of Medical Informatics and Knowledge Engineering in e-Health) and Kereval. Its objective is to develop and evaluate innovative targeting algorithms to identify patients eligible for clinical trials, based on hospital data sets. The use cases will focus on two major disease areas: head and neck cancers and colorectal cancers.

Methodology

The methodology is structured around three components:

• Defining a minimum dataset based on both structured and unstructured data from the AP-HP health data warehouse.
• Developing natural language processing (NLP) algorithms to extract key clinical variables, along with targeting algorithms to pre-select patients eligible for clinical trials.
• Evaluating the approach using a protocol that integrates standard metrics and advanced trustworthy AI criteria (robustness through metamorphic testing, and FAIR data principles—Findable, Accessible, Interoperable, Reusable), tested under real-world conditions.

Kereval’s Role

In this project, Kereval, drawing on its expertise in testing Artificial Intelligence (AI)-based systems, will be responsible for evaluating and testing NLP algorithms and targeting algorithms. Furthermore, given its strong involvement in interoperability issues, its teams will work on data FAIRification. 

A French and European Challenge

This two-year project has been selected as part of a national call for expressions of interest led by the Health Data Hub, with the goal of contributing to the Open Library of Health Algorithms (BOAS).

The project will result in the development of algorithms made available to the scientific community as open source, enabling the extraction of information and the identification of patients eligible for clinical trials. It will also provide a comprehensive evaluation protocol (standard metrics and advanced indicators) that can be reused in other AI healthcare projects. Finally, detailed documentation (scope, performance, and conditions of use) will be published as open-access to facilitate reuse of the developed tools.

This initiative aligns with a key strategic focus for Kereval, combining dual expertise in interoperability and artificial intelligence. These capabilities are also being applied in European projects such as i2X and myHealth@myHands, supported by the European Commission. These topics are central to French and European strategies on AI and health data, particularly with the development of the European Health Data Space (EHDS).

Watch the project presentation video