State-of-the-Art · Clinical Cardiology AI

Cardiology guidance,
grounded in ESC evidence.
Covering  

CorGuide is a specialised AI system for clinical cardiology, achieving 95.1% accuracy on clinical cardiology examinations — the highest documented performance of any AI system evaluated in this domain. Built on a proprietary algorithm that reasons across the full spectrum of cardiovascular medicine, with citations of class of recommendation and level of evidence in every answer.

Open clinical chat
Clinical Validation
95.1%
accuracy on cardiology exam · n = 119 cases
CorGuide AI 95.1%
ESC Chatbot 65.7%
ChatGPT 48.4%
Standardised clinical cardiology examination
n = 119 cases
—
Clinical Accuracy
—
vs ChatGPT
—
ESC Guideline Areas
—
Cases Evaluated
Coverage
Full-spectrum cardiovascular medicine
CorGuide's algorithm covers the complete breadth of clinical cardiology — from acute presentations and complex clinical cases to chronic disease management and medical report interpretation — continuously aligned with the most current evidence and guideline directions.
  • Heart Failure (HFrEF, HFmrEF, HFpEF)
  • Cardiovascular Disease Prevention
  • Hypertension Management
  • Endocarditis
  • Ventricular Arrhythmias & Sudden Cardiac Death
  • Cardiac Pacing & CRT
  • Atrial Fibrillation
  • Dyslipidaemias
  • Chronic Coronary Syndromes
  • Acute Coronary Syndromes
  • Valvular Heart Disease
  • Diabetes & Cardiovascular Disease
  • Cardiovascular Disease in Oncology
  • + 11 additional clinical areas
Methodology
How CorGuide produces its answers
A two-stage pipeline that reasons over the question before retrieving guideline context — minimising hallucination and maximising clinical relevance.
01

Query Reasoning

Extended-thinking AI identifies the most relevant ESC guideline chapters for the clinical question, without retrieving any content yet.

02

Precision Retrieval

The algorithm retrieves the most clinically relevant knowledge for the specific question, optimised for accuracy and completeness across all covered areas.

03

Grounded Synthesis

An answer is composed from the retrieved context, citing class of recommendation (I–III) and level of evidence (A–C) directly.

Clinical Validation
Performance on standardised cardiology examination
Evaluated on 119 clinical cases from a standardised cardiology examination — achieving 95.1% accuracy, outperforming both the official ESC Chatbot (65.7%) and ChatGPT (48.4%).
Model
Score
vs. Best
CorGuide AI
RAG + Extended Reasoning · ESC-specific
95.1%
—
ESC Chatbot
Official ESC guideline chatbot
65.7%
−29.4 pp
ChatGPT
General-purpose language model
48.4%
−46.7 pp

Standardised clinical cardiology examination · n = 119 cases

In Practice
Evidence-referenced answers in seconds
Every response includes the class of recommendation and level of evidence, enabling rapid clinical reference and verification.
Clinician
What are the first-line treatments for HFrEF according to current guidelines?
CorGuide
According to the 2021 ESC Heart Failure Guidelines, four drug classes carry a Class I, Level A recommendation for HFrEF (LVEF ≤ 40%):
  • ACE inhibitor (or ARB/ARNI if not tolerated) — Class I, Level A
  • Beta-blocker (bisoprolol, carvedilol, metoprolol succinate, nebivolol) — Class I, Level A
  • Mineralocorticoid receptor antagonist (spironolactone, eplerenone) — Class I, Level A
  • SGLT2 inhibitor (dapagliflozin or empagliflozin) — Class I, Level A
Sacubitril/valsartan is recommended to replace ACE-I in ambulatory patients remaining symptomatic on optimal therapy (Class I, Level B). ESC 2021 Heart Failure §7.3.2, p. 3575.
Capabilities
Designed for clinical use

Current Evidence

Answers grounded in the most up-to-date cardiology knowledge — the algorithm is continuously aligned with the latest clinical guidelines.

Class & Level of Evidence

Every recommendation includes its COR (I–III) and LOE (A–C) as specified in the source guideline.

Multi-Turn Context

Full conversational memory within a session — follow-up questions are answered with awareness of prior exchanges.

Source Attribution

Each response cites the specific guideline, section number, and page — enabling rapid verification at the source.

Ready to use CorGuide?

Evidence-based cardiology answers at the point of clinical need.

Open clinical chat