In order to bring you the best possible user experience, this site uses Javascript. If you are seeing this message, it is likely that the Javascript option in your browser is disabled. For optimal viewing of this site, please ensure that Javascript is enabled for your browser.

e-Computational cardiology – new ways that technology can improve prediction in coronary syndromes

27 Aug 2021

Artificial intelligence (AI) and other innovative computational techniques have tremendous scope to transform cardiology. At ESC Congress 2021, we have seen many examples in the e-Health/digital health track – here are some e-Poster highlights on advances in acute/chronic coronary syndromes.

Doctor Adam Bakula (University Hospital Basel, Switzerland and Royal Brompton Hospital, London, UK) and colleagues investigated whether an AI-based approach, using a memetic pattern-based algorithm (MPA), may be able to improve ischaemia prediction beyond known pre-test probability (PTP) scores.

Over 500 consecutive patients undergoing rubidium-82 positron emission tomography myocardial perfusion imaging (PET MPI) for routine clinical evaluation of coronary artery disease (CAD) were included in the analysis. The PTP for each patient was estimated using the MPA, by using ESC models from 2013 and 2019, by Diamond and Forrester scores (DFS) and by Framingham scores (FRS). PET MPI images were assessed for the presence of ischaemia, defined as a summed difference score (SDS) >=2.

Of the 531 patients included, 50% had known prior CAD and 208 patients had evidence of ischaemia. The MPA provided an area under the receiver operating characteristic (AUC ROC) curve of 0.76, which was higher than with ESC models (ESC 2013 and ESC 2019, both 0.67), DFS (0.56) and FRS (0.68).

It was concluded that MPA outperforms established scores for PTP assessment and has the potential to improve the accuracy of ischaemia prediction, which could help avoid unnecessary testing and save costs.

The use of biomarkers in combination with clinical variables is a promising novel noninvasive approach to predicting obstructive CAD (oCAD). Assistant Professor Johannes Neumann (University Heart Center Hamburg, Clinic Cardiology, Germany) describes how a clinical/proteomic biomarker panel was previously developed using proteomics and AI, and included high­sensitivity cardiac troponin­I (hs-­cTnI), since hs-cTnI is frequently measured in symptomatic patients.1 The panel consists of three clinical variables (male sex, age and previous percutaneous coronary intervention) and three biomarkers (hs-cTnI, adiponectin and kidney injury molecule-1).

What is new is the validation of this model in 924 patients with a mixture of acute and lesser acute presentations from three cohorts. oCAD was defined as >50% coronary obstruction in at least one coronary artery in one cohort or >70% coronary obstruction in at least one coronary artery for the other two cohorts.

Neumann et al. found that the panel had an AUC ROC curve of 0.80 (95% confidence interval [CI] 0.77 to 0.83; p<0.001) for the presence of oCAD. At the optimal cut-off, sensitivity was 74%, specificity was 72%, and the positive predictive value (PPV) was 81% for oCAD. The panel had a diagnostic odds ratio of 7.48 (95% CI 5.55 to 10.09; p<0.001). In comparison, hs­cTnI alone had an AUC ROC curve of 0.63 (95% CI 0.60 to 0.67; p<0.001), with 49% sensitivity, 72% specificity and a positive predictive value of 74%.

The authors concluded that this clinical/biomarker‐based diagnostic model can predict the presence of oCAD with high accuracy across a multinational pooled cohort.

Virtual fractional flow reserve (vFFR) utilises computational fluid dynamics to compute haemodynamics on a reconstructed patient-specific coronary artery model and is being evaluated as an alternative to invasive functional assessment (iFA).
Doctor Tania Mano (Hospital de Santa Marta, Lisbon, Portugal) describes a retrospective analysis to investigate the difference in vFFR analysis between vessels and specific lesions. In 95 consecutive patients who underwent iFA, vFFR was calculated based on coronary angiograms acquired in at least two different projections. Diagnostic performance of vFFR was evaluated and correlated with iFA, according to coronary vessel, vessel diameter at stenosis, diameter stenosis and area of stenosis at the lesion. vFFR <0.80, FFR <0.8 and resting full-cycle ratio (RFR) <0.90 were considered positive. Indications by invasive coronary angiography were chronic coronary syndrome in 63% of patients or acute coronary syndrome (non­culprit lesion) in the remaining patients.

vFFR accuracy was found to be good when vFFR was measured in the distal vessel segment (AUC 0.84, p<0.001; Pearson’s correlation coefficient r=0.53; p<0.001). The correlation improved when vFFR was assessed at the lesion site (r=0.63; p<0.001) or up to 1 cm below the stenosis (r=0.61; p<0.001).

Binary concordance was 89% in the right coronary artery and left anterior descending artery (sensibility 68%, specificity 96%, PPV 87%) and 77% in the circumflex coronary artery (sensibility 50%, specificity 82%, PPV 33%). Correlation between vFFR and iFA was high when vessel diameter at stenosis was ≥2 mm (r=0.73; p<0.001), when diameter of stenosis was <30% (r=0.72; p<0.001), and in areas of mild stenosis (<1.0 mm²; r=0.83; p<0.001) or severe stenosis (>6.0 mm²; r=0.84; p<0.001).

Thus, vFFR appears to have moderate-to-high linear correlation to iFA, depending on the artery and type of lesion studied – information on situations providing the highest correlation are helpful for the next steps in vFFR validation.


For further details, check out the e-Posters at:
Bakula A, et al
Neumann J, et al
Mano T, et al

And follow the Great Debate: Artificial intelligence in cardiology – a marriage made in heaven or hell?

Want to know more about how cutting-edge technology may reshape the management of cardiovascular diseases? The ESC Digital Summit is taking place online from Friday, 22 October to Sunday, 24 October 2021. From AI to robotics, to big data, apps and wearables – all of these topics will be covered. Learn more
The scientific programme is live - explore it.


1. McCarthy CP, et al. J Am Heart Assoc. 2020;9:e017221.