Article Impact Level: HIGH Data Quality: STRONG Summary of European Heart Journal - Cardiovascular Imaging, jead246. https://doi.org/10.1093/ehjci/jead246 Dr. Eero Lehtonen et al.
Points
- The research aimed to evaluate the additional predictive value of positron emission tomography (PET) myocardial perfusion imaging (MPI) over coronary computed tomography angiography (CCTA) for short- and long-term outcomes in patients with coronary artery disease (CAD) using machine learning (ML) techniques.
- A total of 2411 patients suspected of CAD underwent both CCTA and PET MPI evaluation, and two sets of ML models were trained: one incorporating clinical and imaging data (including PET) and the other using clinical and CCTA data alone.
- After removing incomplete data, 2284 patients were included in the analysis, and an 8-year follow-up revealed 210 adverse events, including myocardial infarctions, unstable angina pectoris, and deaths.
- The results demonstrated that PET MPI significantly improved outcome prediction over CCTA during the first four years of observation, with the highest area under the receiver operating characteristic curve (AUC) occurring at Year 1 (AUC: 0.82).
- Beyond the initial 4-year period, PET MPI did not provide significant additional predictive value, highlighting the complementary nature of anatomic and functional information in predicting outcomes in patients with suspected CAD.
Summary
In a study aimed at assessing the added value of positron emission tomography (PET) myocardial perfusion imaging (MPI) in comparison to coronary computed tomography angiography (CCTA) for predicting short- and long-term outcomes in patients with coronary artery disease (CAD), machine learning (ML) techniques were employed. The research, encompassing 2411 patients with clinical indications of CAD, involved the initial performance of CCTA, followed by downstream PET MPI to evaluate hemodynamic aspects of obstructive coronary stenosis in 891 patients. Two sets of Extreme Gradient Boosting (XGBoost) ML models were developed: one incorporating clinical and imaging data, including PET, and the other using solely clinical and CCTA-derived data. After filtering out incomplete data entries, 2284 patients were retained for further analysis.
Over an 8-year follow-up period, the study identified a total of 210 adverse events, which included 59 cases of myocardial infarctions, 35 instances of unstable angina pectoris, and 116 deaths. The results showed that PET MPI data significantly improved outcome prediction when compared to CCTA during the initial four years of observation, with the highest area under the receiver operating characteristic curve (AUC) occurring at Year 1 (AUC: 0.82, 95% confidence interval 0.804–0.827). Subsequently, beyond these four years, the PET MPI did not offer any notable incremental prognostic value.
In conclusion, the study findings suggest that PET MPI variables enhance the prediction of adverse events in patients suspected of CAD, particularly during the first four years of follow-up. This underscores the complementary nature of combining anatomic and functional information in forecasting patient outcomes, demonstrating the potential benefits of utilizing PET MPI alongside CCTA in clinical decision-making.
Link to the article: https://academic.oup.com/ehjcimaging/advance-article/doi/10.1093/ehjci/jead246/7286588
References Lehtonen, E., Kujala, I., Tamminen, J., Maaniitty, T., Saraste, A., Teuho, J., Knuuti, J., & Klén, R. (2023). Incremental prognostic value of downstream positron emission tomography perfusion imaging after coronary computed tomography angiography: A study using machine learning. European Heart Journal - Cardiovascular Imaging, jead246. https://doi.org/10.1093/ehjci/jead246
