AI in Cardiology

This website, "AI in Cardiology," was designed to reflect the contents of a PowerPoint presentation by Dr. Sirin Apiyasawat, MD, on the applications of artificial intelligence in cardiology. The website is structured into sections, each corresponding to a slide in the presentation. The coding was generated by ChatGpt 4.0.

Modern AI: What Made It So Effective?

Large-scale datasets contribute significantly to modern AI's development, allowing for advanced capabilities in recognition and analysis.

Fei-Fei Li, PhD - “The Godmother of AI”

Examples of AI in Cardiology

• autoEF: Comparable to human measurement, with less variance in repeated tests. View on GitHub
• AI in Stress Echo: PROTEUS trial results suggest benefits in low-volume centers.
• Heat Map for MR Visualization: Validated with 80% accuracy in MR quantification (Long et al. 2024).

AI in Mitral Regurgitation

AI quantifies MR with 80% accuracy using color Doppler models. High accuracy achieved when multiple views are used.

Sources: Long et al. & Vrudhula et al., Circulation, 2024

AI-based Quantitative Coronary Angiography (AI-QCA)

RCTs showed similar outcomes between AI-QCA and human assessments, with higher stent malapposition in the AI group.

Source: Kim et al., JACC Intv., 2024

AI-FFR: Fractional Flow Reserve

The PROVISION study demonstrated similar revascularization rates with significant cost reductions and lower radiation in AI-guided groups.

Source: PROVISION Study, TCT 2024

AI-ECG Interpretation

Studies show AI can interpret ECGs with a 2% misinterpretation rate, with atrial fibrillation (AF) as the most common misdiagnosis.

Source: J Cardiovasc Electrophysiol, 2023

ECG-AI Mortality Prediction

AI-ECG Alert reduced 90-day mortality in high-risk groups. AI-based risk stratification effectively identifies asymptomatic left ventricular systolic dysfunction (EF ≤ 35%).

Source: Lin et al., Nat Med, 2024

AI Using Photoplethysmography (PPG) for AF Detection

AI can identify atrial fibrillation using PPG technology from wearables, followed by confirmatory ECG diagnosis.

Source: 2024 ESC Guidelines for AF

AI for Electronic Medical Records (EMR)

AI assists in predicting mortality, readmission, and length of stay, providing insights into patient data and suggesting diagnoses.

Source: NPJ Digit Med, 2018

The Dark Side of AI

Challenges include information overload, "infoxication," alert fatigue, and loss of the human touch, as well as risks of deepfake content.

Sources: Scientific American & JAMA Cardiol, 2021

Best Practices for AI in Cardiology

• Do's: Ensure data quality, use AI to augment decisions, encourage explainable AI, and stay updated.
• Don'ts: Avoid over-reliance on AI, black-box models for critical diagnoses, and overhyping AI to patients.

Quote: "AI won't replace humans, but humans using AI will." - Fei-Fei Li