AI in medicine: from imaging to assistants, what progress is being made?
Imaging devices, multimodal assistants and drug research: understanding practical advances in medical AI and the limits of each use.
Artificial intelligence in medicine does not describe a single tool. Software that helps analyse an image, an assistant that summarises a document and a method used in a clinical trial meet different needs. Understanding progress means looking at the task performed and the evidence behind it.
Imaging: established uses with defined scope
The FDA list identifies AI-enabled medical devices authorised for marketing in the United States. It includes tools in radiology, ultrasound and cardiology. Medical AI has therefore moved beyond laboratory demonstrations.
However, the list concerns specific products and uses. It is not exhaustive, does not validate AI tools in general and does not constitute European authorisation.
Multimodal assistants expand the possibilities
Models that can process text and images open another field. In its 2024 guidance, the WHO describes possible applications in documenting consultations, training healthcare workers and research.
This potential does not guarantee reliable answers. WHO highlights errors, bias and the risk of trusting a machine too readily. Summarising a medical report may help organise information; that alone does not establish that a tool can diagnose a condition. Its intended context and human verification remain essential.
Drug research is also reaching milestones
In March 2025, EMA qualified a methodology using AIM-NASH to help a pathologist assess liver biopsies in clinical trials. This recognition concerns a method for generating evidence, rather than authorisation of a treatment.
The same page presents joint EMA–FDA principles published in January 2026 for AI use in drug development. Progress therefore also concerns how these tools are assessed and monitored.
Distinguishing technical progress from demonstrated benefit
Three questions help assess an announcement: what task does the tool perform, in what setting was it evaluated, and who checks its output? An impressive answer or a well-analysed image does not, by itself, demonstrate improved care.
AI is changing the tools available. Understanding their precise role makes it easier to appreciate progress while distinguishing assistance, experimentation and medical decisions.
Sources
- OMS — Modèles multimodaux en santé : usages, risques et gouvernance — accessed 2026-10-28
- FDA — Liste des dispositifs médicaux intégrant l’intelligence artificielle — accessed 2026-10-28
- EMA — Intelligence artificielle et cycle de vie des médicaments — accessed 2026-10-28
Read also
- Agents IA & automatisationNovember 6, 2026
Agentic RAG: when to turn retrieval into a tool an agent can call
A standard RAG follows a fixed sequence. As soon as multi-step reasoning or source selection is needed, an agent that invokes search as a tool becomes relevant.
Read article - Réseaux sociaux pour TPENovember 6, 2026
Social media for small businesses: selling without becoming a creator
Instagram, TikTok, Facebook: how a small business can use social media to sell, how often to post, and how to measure what actually brings in customers.
Read article - Livraison localeNovember 5, 2026
Livraison locale : évaluer les plateformes sans perdre la marge
Stuart, Uber Eats ou coursiers indépendants : comment comparer les options de livraison locale sans se fier aux promesses. Une approche pragmatique pour les commerçants qui veulent mesurer les coûts réels et décider entre interne et externe.
Read article
