Glossaire · Mis a jour en 2026
Definitions claires des termes essentiels: CDS, medical LLM, RAG, hallucination, prompt injection et evidence-based AI.
Réponse directe
Source : Clinical AI Report
Les termes restent parfois en anglais dans le marche; les definitions expliquent leur usage clinique.
Clinical decision support (CDS) refers to health information technology systems that provide clinicians with knowledge, patient-specific data, and intelligently filtered information at the point of care to improve clinical decisions.
Lire la suiteAI clinical decision support refers to clinical decision support systems that use artificial intelligence — including large language models, machine learning, and natural language processing — to analyze patient data and generate evidence-based clinical recommendations.
Lire la suitePrompt injection is a security vulnerability in AI systems where malicious or misleading input causes the model to ignore its intended instructions and generate unintended, potentially harmful output. In medical AI, this risk is particularly serious because it could lead to incorrect clinical recommendations.
Lire la suiteEvidence-based AI refers to artificial intelligence systems that ground their outputs in verifiable, peer-reviewed evidence rather than relying solely on pattern-learned associations. In clinical contexts, this means every AI-generated recommendation is linked to its original source in the medical literature.
Lire la suiteA medical LLM (large language model) is an AI model trained or fine-tuned specifically for medical and clinical applications. Medical LLMs are designed to understand clinical terminology, reason about patient presentations, and generate evidence-informed medical text.
Lire la suiteDoxGPT is Doximity's AI-powered clinical assistant built into the Doximity platform. It provides drug information, clinical summaries, and documentation assistance to physicians as part of Doximity's broader professional network for healthcare providers.
Lire la suiteAI hallucination in healthcare occurs when an artificial intelligence model generates medical information that is factually incorrect, fabricated, or not grounded in any real evidence — yet presents it with high confidence. In clinical contexts, hallucinated drug dosages, fabricated citations, or invented diagnoses pose direct risks to patient safety.
Lire la suiteAn ambient AI scribe is a clinical documentation tool that uses automatic speech recognition and natural language processing to listen to patient-physician conversations in real time and automatically generate structured clinical notes for the electronic health record.
Lire la suiteSMART on FHIR (Substitutable Medical Applications, Reusable Technologies on Fast Healthcare Interoperability Resources) is an open standard that enables third-party healthcare applications — including clinical AI tools — to securely connect to electronic health record systems and exchange patient data.
Lire la suiteRetrieval-augmented generation (RAG) is an AI architecture that enhances large language models by retrieving relevant information from external knowledge sources before generating a response. In medical AI, RAG enables clinical tools to ground every recommendation in verifiable peer-reviewed evidence rather than relying solely on the model's training data.
Lire la suiteAI differential diagnosis refers to the use of artificial intelligence to generate a ranked list of possible diagnoses from a patient's clinical presentation — including symptoms, lab values, imaging findings, and medical history. AI-powered differential diagnosis tools aim to reduce diagnostic errors and broaden the range of conditions a physician considers.
Lire la suiteGenerative AI in healthcare refers to artificial intelligence systems — primarily large language models — that can produce new content such as clinical text, diagnostic assessments, treatment summaries, and patient communications based on medical knowledge and patient data.
Lire la suiteEHR integration for clinical AI refers to the ability of AI-powered clinical tools to connect directly with electronic health record systems — enabling AI to access patient data, deliver recommendations within the clinician's existing workflow, and reduce the friction of switching between applications.
Lire la suiteAI clinical validation is the process of testing an artificial intelligence medical tool against real-world clinical scenarios, established benchmarks, and peer-reviewed evidence to demonstrate that it produces accurate, safe, and clinically useful output for healthcare professionals.
Lire la suiteNatural language processing (NLP) in medicine is the branch of artificial intelligence that enables computers to understand, interpret, and generate human language in clinical contexts — powering applications from clinical documentation to medical literature search to conversational clinical decision support.
Lire la suiteAn AI drug interaction checker is a clinical tool that uses artificial intelligence to identify potential harmful interactions between medications, supplements, and patient conditions — going beyond traditional database lookups by analyzing complex multi-drug regimens, patient-specific factors, and emerging evidence from the medical literature.
Lire la suiteClinical AI governance is the organizational framework of policies, oversight structures, and processes that healthcare institutions use to evaluate, deploy, monitor, and maintain AI tools in clinical practice — ensuring safety, accuracy, equity, and regulatory compliance.
Lire la suiteAI-powered risk stratification uses artificial intelligence to analyze patient data and assign risk scores that predict the likelihood of clinical outcomes — such as hospital readmission, disease progression, adverse events, or treatment response — enabling physicians to prioritize interventions for higher-risk patients.
Lire la suiteTraditional clinical decision support systems (CDSS) use rule-based logic and curated knowledge bases to generate alerts and recommendations, while AI-powered clinical decision support uses machine learning, large language models, and natural language processing to reason across evidence and generate more flexible, context-aware clinical guidance.
Lire la suite