AI in the Cath Lab: Revolution with Caution

A quiet revolution is unfolding in interventional cardiology. At the 2026 EAPCI Summit in Munich, an entire session was devoted to a single subject: artificial intelligence. From predictive modeling to real-time imaging guidance, AI is no longer a futurist’s fantasy — it is increasingly woven into the everyday fabric of cardiovascular care.

A leading interventional cardiologist from the University of Galway opened with a sweeping overview. AI, he explained, “encompasses any technique or system that allows a computer to mimic human behavior: feeling, thinking, acting, and adapting.” Its forms — from machine learning to large language models — are already being deployed to accelerate diagnoses, predict outcomes, and streamline clinical workflow.

Unexpected Discovery

Using machine learning across 75 preprocedural factors in the SYNTAX trial, researchers uncovered a surprise 10-year mortality predictor: gamma-glutamyl transferase (GGT) — a marker of systemic oxidative stress. Patients with high GGT faced 32.7% mortality vs. 23.5% for those with low GGT.

Beyond mortality prediction, AI is reshaping how cardiologists prepare for procedures. Pre-procedural simulation using coronary CT angiography (CCTA) can model anatomy, blood flow, and stenting outcomes — even rendered in virtual reality for heart team review before a patient ever enters the cath lab. “Planning is everything,” the researcher declared. AI lets clinicians “interrogate just about everything” in advance.

A researcher from Barts Heart Centre, London, focused on intravascular imaging. Intravascular ultrasound (IVUS) and optical coherence tomography (OCT) systems are increasingly AI-augmented, capable of characterizing plaque composition and predicting PCI results in near real-time. “AI has revolutionized intravascular imaging processing,” they said.

---

Yet for all the promise, the Barts researcher issued a firm caution. Most AI solutions for intravascular imaging “still make mistakes as they have been trained in small datasets.” Clinical trust, they stressed, requires validation in rigorous outcomes studies — a bar the field has not yet cleared. AI “cannot replace the cardiologist’s own judgment,” and any tool deployed in humans must be thoroughly audited.

A cardiologist from the Cardiovascular Research Foundation surfaced another compelling argument for AI adoption: reducing variability. Across trial databases, identical diagnoses can be managed very differently in the United States versus Japan. AI-driven standardization, over time, could narrow that gap and strengthen the evidence base for all patients.

The Galway researcher offered a candid glimpse into their own daily workflow: they dictate clinical questions, ChatGPT refines the language, and the system organizes tasks for the following day. “It’s something that is a real tsunami in our practice” — efficiency that ultimately frees physicians for more meaningful face-to-face time with patients.

The session closed with a question posed by a co-moderator from the University Hospital of Pisa: what happens when an AI tool earns a Class I guideline recommendation? The answer remains unwritten. What is clear, however, is that AI will reshape cardiovascular practice — carefully, thoughtfully, and with both eyes open to the shadows that accompany its light.