Machine Learning (ML) is revolutionizing medical data analysis. ML algorithms offer the possibility to evaluate large numbers of interrelated variables to predict CAD-related outcomes. Sequential hybrid imaging with PET/CCTA has demonstrated diagnostic and prognostic value in patients with suspected CAD. However, traditional prognostic analyses may be unable to exploit the full potential the resulting structured clinical, CCTA and PET data. Therefore, we aimed to implement a sequential ML workflow to analyze clinical and hybrid PET/CT data for the identification of patients who developed myocardial infarction (MI) or death in a long-term follow up registry.
Data from 951 symptomatic patients with an intermediate risk of CAD was analyzed. Patients underwent sequential hybrid 15O-water PET/CCTA and were followed for an average of 6 years for the development of MI or cardiac death. Clinical variables were extracted from the electronic patient records (sex, age, smoking, diabetes, hypertension, dyslipidemia, family history, chest complaints, dyspnea and early revascularization). CCTA images were evaluated for system dominance and segmentally for: presence of atherosclerotic plaque, % of luminal stenosis and plaque calcification. Thereon, absolute stress PET myocardial perfusion data was evaluated regionally (ml/g/min). After feature selection, modeling was conducted utilizing a 10-fold cross validated boosted ensembles approach (LogitBoost) to analyze clinical, PET and CCTA data. Predictive performance for the development of MI or death was evaluated through AUCs and accuracy.
There were 525 women (55%). Mean age was 61±9 years. 24 MI and 49 all-cause deaths were documented during follow-up (range: 1 month – 9.6 years), while 109 patients underwent early revascularization. With a total of 85 variables (10 clinical, 58 from CCTA and 17 from PET) analyzed, predictive performance was discrete for clinical data (AUC=0.65,Acc=90%) and moderate for clinical + quantitative PET data (AUC=0.69,Acc=92.5%), while there was significant performance improvement (p=0.005) when integrating clinical + quantitative PET + CTA data (AUC=0.82,Acc=95.4%) in the identification of patients who experienced MI or all-cause death during follow up independently from early revascularization.
Stepwise ML for the analysis of clinical and hybrid sequential cardiac PET/CCTA data can improve the identification of symptomatic patients with suspected CAD who will develop MI or death in long-term follow up. This proposes the added prognostic value of ML in cardiac hybrid imaging.