Generative AI in healthcare: transformations & challenges

Visual identity, video production & graphics: Ines Kedjem

  • Dr. Stéphane Ohayon

Dr. Stéphane Ohayon’s presentation explores the transformations and challenges of generative AI in healthcare.
It reviews the fundamentals of AI, from machine learning and deep learning to generative models.
Generative AI represents a form of “narrow” AI, capable of producing content without true understanding or autonomy.

In healthcare, it supports medical imaging and early anomaly detection.
It also enhances consultations through automatic transcription and report generation.
AI contributes to knowledge synthesis, clinical decision support, and care organization.
The quality of outputs strongly depends on precise and well-structured prompts.
Adoption remains limited due to insufficient professional training and patient mistrust.
Key challenges include data protection, bias mitigation, transparency, and medical responsibility.
Generative AI must operate within a strong ethical and human framework to support, rather than replace, care.

AI & Digital Twins

Visual identity, video production & graphics: Ines Kedjem

  • Portrait de Fabrice Denis, expert in prevention, health and well-being

    Prof. Fabrice Denis

Professor Fabrice Denis’ presentation examines the role of digital twins and AI in preventive, longevity-oriented healthcare.
A digital twin is a virtual model of an individual’s health status and biological aging.
It relies on the analysis of around 600 biological, clinical, and functional parameters.

These data generate multiple biological ages, including phenotypic, metabolic, vascular, and epigenetic.
Combined, they provide an objective dashboard to assess global and organ-specific aging risks.
AI models, particularly large language models (LLMs), could theoretically assist in data extraction and analysis.
However, significant limitations remain, including errors, hallucinations, and lack of reproducibility.
Clinical cases reveal substantial discrepancies between AI estimates and validated medical assessments.
Clinical examination and complementary testing therefore remain essential.
Digital twins show strong promise, but LLMs require strict oversight and further optimization before reliable clinical use.