AI in healthcare, a silent but profound revolution

Imagine a system capable of analysing thousands of X-rays in a few seconds, of detecting a cancer at an early stage invisible to the human eye, or of predicting a cardiac decompensation 48 hours before it occurs. This is not science fiction: it’s artificial intelligence in healthcare, and it is already profoundly transforming medical practices around the world.

Yet for many healthcare professionals, AI remains a vague, even intimidating concept. What is artificial intelligence applied to medicine? How does it actually work? What are its real uses in hospitals and practices today? And above all, what does it mean for your daily practice?

Definition: what is artificial intelligence in healthcare?

Artificial intelligence (AI) refers to the set of computer technologies capable of simulating certain human cognitive functions: learning, reasoning, recognising patterns, understanding natural language or making decisions.

Several major families of AI techniques used in healthcare can generally be distinguished:

• Machine learning consists of training statistical models on large quantities of data so that they can then make predictions or classifications.
• Deep learning is a subcategory of machine learning that uses artificial neural networks inspired by the human brain. It is this technology that lies behind the spectacular progress in medical imaging.
• Natural language processing (NLP) enables machines to understand and generate text. In healthcare, it is used to automatically extract structured information from medical reports.
• Generative AI, whose most visible representatives are the major conversational AIs such as ChatGPT or Claude, opens up new possibilities: summarising patient records, assisting with the writing of reports, answering patients’ questions, or supporting medical training.

A brief history: AI in healthcare was not born yesterday

The idea of using computing to assist diagnosis dates back to the 1970s with the first expert systems such as MYCIN, developed at Stanford to help diagnose bacterial infections and prescribe antibiotics.

• $100bn: Global market for AI in healthcare by 2030
• 2012: Breakthroughs in deep learning for imaging
• 1970: First medical expert systems

In France, the Agence du Numérique en Santé (ANS) and the national AI strategy place healthcare among the priority application areas.

The major application areas of AI in healthcare

1. Medical imaging: the most mature field

Medical imaging is undoubtedly the sector where AI has demonstrated the most convincing results. AI algorithms are today capable of detecting suspicious pulmonary nodules, identifying early signs of diabetic retinopathy, analysing mammograms or segmenting tumours on brain MRIs.

In France, solutions such as Gleamer or Therapixel are already deployed in several hospital establishments. Internationally, the IDx-DR software was the first AI device to obtain marketing authorisation from the American FDA for autonomous diagnosis (diabetic retinopathy screening).

2. Biology and genomics

Genomic sequencing generates considerable volumes of data, impossible to analyse manually in an exhaustive way. AI makes it possible to identify genetic variants associated with rare diseases, to predict a patient’s response to a treatment in oncology (precision medicine), or to detect antibiotic-resistance mutations.

3. Prediction and preventive medicine

One of the most promising applications of AI in healthcare is its ability to predict clinical events before they occur:

• Models capable of predicting the risk of sepsis several hours before the first clinical signs by analysing in real time the vital signs of a patient in intensive care
• Algorithms that identify diabetic patients at high risk of developing kidney failure, making it possible to intensify their follow-up
• Tools that detect in advance patients likely to decompensate after hospital discharge, reducing avoidable readmissions

4. Clinical decision support

AI-based clinical decision support systems (CDSS) integrate directly into electronic patient record (EPR) software. They can alert a prescriber to a drug interaction, suggest a list of differential diagnoses, recommend a HAS (French National Authority for Health) protocol, or check the consistency of a prescription.

5. The patient relationship and communication tools

• Assisted report writing: tools such as Nabla or Suki automatically transcribe a medical consultation and generate a structured report in a few seconds
• Triage chatbots: conversational agents guide patients to the right level of care
• Patient record summaries: AI can summarise in a few lines a patient record of several hundred pages

What AI in healthcare is not: debunking misconceptions

Will AI replace doctors?

The answer is no; at least not within a foreseeable horizon. AI excels at well-defined, reproducible tasks. At this stage it cannot replace global clinical reasoning, empathy, the therapeutic relationship, or the management of novel and complex situations.

Is AI always right?

No. AI algorithms have limits and can make mistakes. A model trained on a given population may underperform on another. This is why human supervision remains indispensable.

Is AI neutral?

Not necessarily. Biases can creep into the training data and be reproduced, or even amplified, by the algorithm. Studies have shown that some dermatological image-recognition systems were less effective on dark skin, because the training data lacked diversity.

The regulatory and ethical framework in France and Europe

• The GDPR imposes strict rules on the collection, processing and storage of health data
• The European Medical Device Regulation (MDR) applies to AI software that has a medical purpose
• The Artificial Intelligence Act (AI Act), which came into force in 2024, classifies most AI in healthcare in the “high risk” category
• The HAS published in 2023 an evaluation framework for AI in healthcare

Why must healthcare professionals train in AI today?

AI in healthcare is no longer a technology of the future: it is already present in many hospital departments, and its spread is going to accelerate in the years to come. Training in AI does not mean becoming a developer or a data scientist. It means acquiring the knowledge needed to understand the tools, evaluate their results, take part in your establishment’s decisions, protect your patients and enhance your practice.

AI in healthcare, an opportunity to seize

Artificial intelligence in healthcare is a complex, multifaceted and rapidly evolving reality. It is neither the panacea that some announce, nor the threat that others fear. It is a set of powerful tools which, well understood and well used, can help improve the quality of care, patient safety and the working conditions of healthcare professionals.

The key lies in the training and upskilling of care providers. Because the most high-performing AI remains useless, or even dangerous, in hands that do not understand it.

Would you like to go further and acquire concrete mastery of AI for your professional practice? Discover our training courses specially designed for healthcare professionals: accessible, progressive and grounded in real use cases from the medical field.

Sources & references

1. Shortliffe, E. H. & Buchanan, B. G. (1975). “A model of inexact reasoning in medicine.” Mathematical Biosciences, 23(3-4), 351–379. [MYCIN system]
2. LeCun, Y., Bengio, Y. & Hinton, G. (2015). “Deep learning.” Nature, 521(7553), 436–444.
3. Statista / Grand View Research (2023). “Artificial Intelligence in Healthcare Market Size — Global forecast to 2030.”
4. Gleamer. Clinical performance studies and CE marking: gleamer.ai.
5. Therapixel. AI in mammography: therapixel.com.
6. Abràmoff, M. D. et al. (2018). “Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy.” npj Digital Medicine, 1, 39. [IDx-DR, first FDA authorisation]
7. Agence du Numérique en Santé (ANS). Digital health roadmap 2023-2027. esante.gouv.fr.
8. Haute Autorité de Santé (HAS). “Evaluation of medical devices incorporating artificial intelligence.” Methodological guide, 2023.
9. Regulation (EU) 2024/1689 — Artificial Intelligence Act. Official Journal of the European Union.
10. Adamson, A. S. & Smith, A. (2018). “Machine learning and health care disparities in dermatology.” JAMA Dermatology, 154(11), 1247–1248.
11. Nabla. Product documentation: nabla.com.
12. Suki. Product documentation: suki.ai.