Artificial intelligence and logistic regression vs. the modified Duke score in the detection of infective endocarditis: a comparison within the PRO-ENDOCARDITIS study

Background: The modified Duke score as current standard in the diagnostic of infective endocarditis (IE) is strictly categorized into major and minor criteria. Therefore, individual patient’s data may not be utilized optimally. In other cardiovascular diseases artificial intelligence approaches outperformed established diagnostic criteria. We tested whether neural networks and logistic regression could improve diagnostic accuracy for IE and calculate a pretest probability prior to transesophageal echocardiography (TEE). Methods: This post-hoc evaluation of the prospective observational PRO-ENDOCARDITIS study includes 261 patients of the West German Heart and Vascular Center enrolled between December 2017 and May 2019. Duke criteria and clinical characteristics were prospectively collected. Positron emission tomography-computed tomography (PET-CT) was performed when clinically indicated. TEE imaging was evaluated by a blinded cardiologist at a central core-lab. IE as primary endpoint was adjudicated by an independent clinical endpoint committee. The performance of the different calculatory approaches was evaluated using the area under the curve (AUC) of the receiver operating curve. Results: In the prediction of endocarditis diagnostic accuracy ranked 1st logistic regression [AUC: 0.984 (0.973, 0.996)], 2nd neural network and 3rd the modified Duke criteria. In the calculation of the pretest probability prior to TEE the logistic regression achieved an AUC of 0.919 (0.879, 0.959). Conclusions: Our results generate the hypothesis that sophisticated statistical analysis and neural nets optimize the prediction of IE and may be utilized to calculate pretest probability prior to TEE.