Artificial intelligence approach improves ischaemia prediction beyond the known pretest probability scores

Abstract

Abstract Background Pre-test probability (PTP) assessment is crucial in the assessment of patients with suspected ischaemia/coronary artery disease. Presently there is no established role for artificial intelligence in estimating PTP. Purpose Comparison of a recently developed memetic pattern based algorithm (MPA) with the diagnostic power of established scores in predicting ischaemia on positron emission tomography myocardial perfusion imaging (PET MPI). Methods Consecutive patients undergoing Rubidium PET MPI for routine clinical evaluation of ischaemia were included. The PTP for each patient was estimated by MPA, the Diamond and Forrester scores (DFS), the PTP models from the European Society of Cardiology Guidelines from 2013 (ESC 2013) and 2019 (ESC 2019) and the Framingham scores (FRS). The PET MPI studies were assessed for the presence of ischaemia. Ischaemia was defined as a summed difference score (SDS) ≥2. Results The mean age of the 531 patients was 66±11 years, 34% were female, and 50% had known prior coronary artery disease; 208 patients had evidence of ischaemia. No ischaemia was found in 323 patients. The areas under the curve (AUC) are shown in the figure. The artificial intelligence based MPA provided an AUC of 0.76, PTP (ESC 2013) AUC of 0.67, PTP (ESC 2019) AUC 0.67, DFS AUC 0.56, FRS 0.68. Conclusion The MPA outperforms the established scores for PTP assessment of the ESC, DFS and FRS in prediction of ischaemia on PET MPI. It has the potential to improve the accuracy of the established diagnostic algorithm for CAD and ischaemia. Funding Acknowledgement Type of funding sources: Public grant(s) – National budget only. Main funding source(s): The study was in part funded by the Swiss Heart Foundation.