Phantomics, in collaboration with Prof. Jin Hur (Department of Radiology, Severance Hospital) and Prof. Jin-Young Kim (Keimyung University Dongsan Hospital), has developed a deep-learning model that not only detects coronary artery disease (CAD) on CT angiography but also predicts future cardiac events.
Method: Deep-learning network automatically detects coronary stenosis and classifies it into three groups—normal, non-obstructive (< 50 % narrowing), and obstructive (≥ 50 %).
Outcomes: During an average 2.5-year follow-up, obstructive patients showed a 38.8 % incidence of major cardiac events vs. 0.6 % in normals.
Adding the AI-derived obstructive status to traditional risk factors improved prognostic accuracy by 14 %.
Impact: Provides rapid CAD diagnosis and long-term risk stratification at the point of care, streamlining decision-making in the emergency setting.
Phantomics, in collaboration with Prof. Jin Hur (Department of Radiology, Severance Hospital) and Prof. Jin-Young Kim (Keimyung University Dongsan Hospital), has developed a deep-learning model that not only detects coronary artery disease (CAD) on CT angiography but also predicts future cardiac events.
Dataset: 408 emergency-department patients (three tertiary hospitals, 2018–2022)
Method: Deep-learning network automatically detects coronary stenosis and classifies it into three groups—normal, non-obstructive (< 50 % narrowing), and obstructive (≥ 50 %).
Outcomes: During an average 2.5-year follow-up, obstructive patients showed a 38.8 % incidence of major cardiac events vs. 0.6 % in normals.
Adding the AI-derived obstructive status to traditional risk factors improved prognostic accuracy by 14 %.
Impact: Provides rapid CAD diagnosis and long-term risk stratification at the point of care, streamlining decision-making in the emergency setting.
Publication: Radiology: Artificial Intelligence (May 19, 2025).
This work demonstrates how AI can evolve beyond image interpretation into a comprehensive clinical decision-support tool.
Related Paper: Predicting Major Adverse Cardiac Events Using Deep Learning–based Coronary Artery Disease Analysis at CT Angiography