REGISTER
ESC 2026 Opens in Munich: Lüscher and AI on the Cardiologist of Tomorrow
ESC 2026 opened in Munich with the theme "Transforming Cardiology, Together" and AI as the inaugural focus. ESC President Prof. Thomas F. Lüscher delivered a keynote titled "The Cardiologist of Tomorrow" alongside an AI moderator. Mayo Clinic AI-ECG for atrial fibrillation (AUC 0.9) and a Taiwan RCT showing AI-assisted ECG reduced 30-day mortality by 30% in high-risk patients (p=0.03) were highlighted. GRACE 3.0 was published in The Lancet. Lüscher warned of cognitive surrender.

Article Key Points

- ESC 2026 opened in Munich with the theme “Transforming Cardiology, Together,” establishing AI as the inaugural focal topic; ESC President Prof. Thomas F. Lüscher delivered a keynote titled “The Cardiologist of Tomorrow,” co-presenting with an AI moderator driven by a large language model.

- AI applications highlighted in cardiovascular medicine include Google’s AMIE system for diagnostic dialogue, voice analysis for pulmonary congestion in heart failure, Mayo Clinic AI-ECG algorithms (AUC 0.9 for prior atrial fibrillation detection, multi-year prediction of LV systolic dysfunction), and the Oxford FAI for pericoronary inflammation.

- A Taiwan RCT (>10,000 emergency patients) published in The Lancet Digital Health showed AI-assisted ECG interpretation reduced 30-day all-cause mortality by 30% in high-risk patients (p=0.03), with no significant difference in low- and intermediate-risk groups.

- A NEJM study reported that GPT-4 significantly outperformed most physicians and nearly matched top experts in complex clinical cases; GRACE 3.0 (Lancet) used machine learning for NSTE-ACS prognosis stratification with consistent performance in both sexes.

- Prof. Lüscher warned of “cognitive surrender” (over-trust in algorithms eroding clinical judgment) and AI hallucinations (fabricated references), and cited a colonoscopy study showing that endoscopists who had used AI performed worse than AI-naive controls after AI was withdrawn.

In August, the Munich exhibition center was packed with attendees. One of the world’s largest and most influential academic congresses in cardiovascular medicine, the European Society of Cardiology (ESC) Congress, formally opened here. This year’s congress is themed “Transforming Cardiology, Together,” with artificial intelligence (AI) for the first time established as the annual focal topic. This signals that the digital transformation of cardiovascular medicine is moving from “possible” to “necessary.”

At the opening ceremony, ESC President Prof. Thomas F. Lüscher and a special “partner,” an AI moderator driven by a large language model, completed the opening interaction. The AI moderator addressed attendees from around the world with fluent, natural language, and as it took center stage on the world’s leading cardiovascular academic platform to participate in hosting, what it foreshadowed was an in-depth conversation about the future. Prof. Lüscher then delivered a keynote titled “The Cardiologist of Tomorrow,” directly addressing the opportunities and challenges of cardiovascular medicine in the AI era.

Fig. 1. Prof. Thomas F. Lüscher and the AI Moderator

Fig. 2. Prof. Lüscher Delivering His Keynote

AI as “Co-Pilot”: Comprehensive Empowerment From Diagnosis to Decision-Making

From the gradual replacement of some manual labor by the steam engine in the 18th century to today’s rapid development of AI technology, medicine is undergoing a shift from “replacing muscle” to “augmenting mind.” Prof. Lüscher cited several studies demonstrating AI’s powerful potential in cardiovascular medicine.

In consultation, Google’s AMIE system can conduct diagnostic reasoning through patient-clinician dialogue; in one randomized controlled trial, AMIE outperformed general practitioners on several evaluation metrics. In voice analysis, an Israeli study showed that AI can assess the degree of pulmonary congestion by analyzing five sentences spoken by patients with heart failure; the proposed mechanism is that pulmonary congestion and associated hemodynamic changes may affect vocal cord function and speech characteristics.

In electrocardiography (ECG), an AI-ECG algorithm developed by the Mayo Clinic team in the United States can identify prior atrial fibrillation during sinus rhythm, with an area under the receiver operating characteristic curve (AUC) of 0.9, approaching the ideal level for a diagnostic test; the same team developed another algorithm that can predict left ventricular systolic dysfunction years in advance. More significantly, they have moved out of the laboratory: a randomized controlled trial published in The Lancet Digital Health, enrolling more than 10,000 emergency patients in Taiwan, showed that when clinicians interpreted ECGs with AI assistance, 30-day all-cause mortality in high-risk patients fell by 30% (p=0.03), while no significant difference was seen in the low- and intermediate-risk groups, suggesting that the benefit of AI is concentrated in high-risk populations.

In imaging, the Fat Attenuation Index (FAI) proposed by an Oxford University team measures attenuation changes in pericoronary adipose tissue on CT and may, to some extent, reflect plaque-related inflammation. In retrospective cohort validation involving thousands of patients, this index showed independent predictive value for long-term cardiovascular mortality.

Can AI Outperform Human Doctors? Data Provides the Answer

A study published in NEJM showed that in the diagnosis of complex clinical cases, generative pre-trained transformer 4 (GPT-4) significantly outperformed most physicians and nearly matched top experts. In addition, Prof. Lüscher’s team used a large language model to integrate multidimensional inputs such as heart failure symptoms, echocardiography, and biomarkers, and found that AI’s diagnostic sensitivity for HFpEF far exceeded that of routine clinical practice.

To this end, Prof. Lüscher announced that the ESC has launched the ESC Chat system, integrating approximately 20 current guidelines, allowing clinicians to ask questions in real time during outpatient consultations and receive evidence-based recommendations.

In risk prediction, results from Prof. Lüscher’s team on the GRACE 3.0 model, published in The Lancet, showed that the model used machine learning to achieve precise prognostic stratification in patients with non-ST-elevation acute coronary syndromes (NSTE-ACS) and performed well in both men and women, addressing the limitations of traditional scores in gender applicability. In addition, AI can perform precise phenotype stratification of patients: for example, MI patients with and without cancer show significantly different clustering across multidimensional parameters such as inflammatory markers, coagulation status, and nutritional status, and the benefit-risk assessment of their optimal intervention strategies (such as conservative versus early invasive treatment, standard versus intensified antithrombotic therapy) differs accordingly. This kind of individualized insight that goes beyond “averages” is the core path by which precision medicine can be realized in the AI era.

Beware “Cognitive Surrender”: Hidden Risks of the AI Era

Prof. Lüscher issued a warning in his keynote: if clinicians over-trust algorithms, they may gradually erode their own clinical judgment and fall into “cognitive surrender.” Studies have shown that when participants are confronted with an algorithm that is intentionally set to be incorrect, their independent judgment performance is actually worse than that of control groups without algorithmic assistance; the tendency to blindly trust AI overwhelms rational questioning. In addition, the phenomenon of AI hallucinations cannot be ignored: algorithms may generate references in proper format that do not actually exist, and for clinicians who lack the habit of critical examination, this constitutes a potential cognitive trap. Studies have shown that AI-assisted colonoscopy improves endoscopists’ polyp detection rate; however, once AI assistance is withdrawn, these clinicians’ independent detection ability becomes inferior to that of control groups who have never used AI. This phenomenon suggests that long-term reliance on AI may lead to skill degradation.

Conclusion

At the end of his keynote, Prof. Lüscher brought his focus back from technology itself to “people.” He concluded that clinical reasoning matters more than memorization. AI can retrieve knowledge in seconds, but the ability to identify problems and question answers always belongs to humans; data and digital literacy means being able to use algorithms critically, not passively accepting them; ethics and privacy awareness are non-negotiable bottom lines; and the human qualities of empathy and trust, together with the time left for patients when AI takes over transactional tasks, are precisely the warmth that machines can never replace.

He closed the session with a maxim from Hippocrates more than two thousand years ago: “It is more important to know the person who has the disease than the disease the person has.” In the AI era, this sentence has acquired entirely new practical significance: when algorithms can efficiently process the data of “diseases,”clinicians can truly focus on the “person” themselves. Personalized medicine is no longer a distant vision, but a reality being gradually fulfilled in this evolution of human-machine collaboration.

0
0
Loading