Fig. 3: Examples of PET/CT imaging in predicted CAD patients.
From: Multicenter evaluation of interpretable AI for coronary artery disease diagnosis from PET biomarkers
Examples of patients undergoing PET/CT myocardial perfusion imaging predicted to have CAD. Bar plots demonstrate the top 5 parameters with the highest influence on per-patient prediction of CAD. a Case where patient has abnormal MFR and stress MBF, borderline stress TPD, zero CAC, left main ≥50% and ≥70% in other three arteries (99% in LAD, 70% in distal LCX, and 70% in distal RCA); AI model estimated the likelihood of CAD to be high in agreement with the clinical score. b Case where patient has normal MFR and stress MBF, borderline ischemic and stress TPD, zero CAC, no actual CAD (20% stenosis in mid RCA; no stenosis elsewhere). The clinical score indicated an abnormal scan. The AI model correctly estimated the likelihood of CAD to be low, aligning with the patient’s actual condition. While iTPD for the patient is borderline, the presence of other factors, such as normal blood flow, reduces the probability of the disease. BMI body mass index, CAC coronary calcium score, CAD coronary artery disease, CT computed tomography, LCX left circumflex artery, LAD left anterior descending, LM left main, LVEF left ventricular ejection fraction, MBF myocardial blood flow, MFR myocardial flow reserve, PET positron emission tomography, RPP rate pressure product, TID transient ischemic dilation, TPD total perfusion deficit, VD vessel disease.
