On 28 March 2023, the HARPOCRATES project was represented at the 5th International Workshop on Health Data Management in the Era of AI (HeDAI 2023), held in Ioannina, Greece. The workshop focused on methods and technologies for managing sensitive health data in AI-driven research and clinical applications.
Antonis Michalas, from Tampere University, presented the HARPOCRATES paper titled “Split Ways: Privacy-Preserving Training of Encrypted Data Using Split Learning.” The presentation introduced a technique developed within the project that enables machine learning model training on encrypted data without the need to decrypt it or share raw inputs between institutions.
This method, known as split learning, allows different parts of the model to be trained across multiple parties while maintaining data confidentiality. The approach directly supports HARPOCRATES objectives in enabling secure collaboration on sensitive medical data, especially in scenarios like the sleep medicine analytics demonstrator, where hospitals need to collaborate without compromising patient privacy.
The HeDAI 2023 workshop provided an academic setting for sharing new ideas and research in privacy-preserving AI for healthcare, aligning closely with the goals of HARPOCRATES. By contributing to the workshop, the project strengthened its scientific impact and engaged with a wider community working on ethical and secure use of health data.
More information about the workshop and full program is available on the HeDAI 2023 website.