Prehospital diagnostic algorithm for acute coronary syndrome using machine learning: a prospective observational study

Abstract

Rapid and precise prehospital recognition of acute coronary syndrome (ACS) is key to improving clinical outcomes. The aim of this study was to investigate a predictive power for predicting ACS using the machine learning-based prehospital algorithm. We conducted a multicenter observational prospective study that included 10 participating facilities in an urban area of Japan. The data from consecutive adult patients, identified by emergency medical service personnel with suspected ACS, were analyzed. In this study, we used nested cross-validation to evaluate the predictive performance of the model. The primary outcomes were binary classification models for ACS prediction based on the nine machine learning algorithms. The voting classifier model for ACS using 43 features had the highest area under the receiver operating curve (AUC) (0.861 [95% CI 0.775–0.832]) in the test score. After validating the accuracy of the model using the external cohort, we repeated the analysis with a limited number of selected features. The performance of the algorithms using 17 features remained high AUC (voting classifier, 0.864 [95% CI 0.830–0.898], support vector machine (radial basis function), 0.864 [95% CI 0.829–0.887]) in the test score. We found that the machine learning-based prehospital algorithms showed a high predictive power for predicting ACS.

Preclinical identification of acute coronary syndrome without high sensitivity troponin assays using machine learning algorithms

“…5 , the feature ST-Elevation as assessed by the emergency physician still is paramount for predicting both endpoints in our models. Not surprisingly, similar findings have been reported by Takeda et al 21 .…”

“…Especially the lack of a prospective validation cohort is the biggest limitation of our study and further analysis is needed. To our knowledge, the only comparable study which used prospectively recorded data was carried out by Takeda et al and achieved slightly better AUROC for the endpoint ACS then our study did for MACE and ACAO (0.86 versus 0.78 and 0.81 respectively) 21 . However, because of the different preclinical emergency systems in Japan and Germany (paramedics versus emergency medicine physician), the studies are only partially comparable.…”

“…To our knowledge, there have been two studies which used machine learning models to predict ACS in a purely preclinical setting. Al-Zaiti et al tried to predict ACS only using data from a preclinical 12-lead ECG whereas Takeda et al used vital signs, history and a 3-lead ECG to predict ACS and myocardial infarction 21 , 22 . Our approach is novel and different in that we chose a different secondary endpoint.…”

International evaluation of an artificial intelligence–powered electrocardiogram model detecting acute coronary occlusion myocardial infarction

“…Several previous studies deployed machine learning to triage patients presenting with ACS, however, bearing multiple limitations. 23 , 31–43 The majority of these studies did not validate the occlusive or flow-limiting culprit lesions on coronary angiogram and relied on a subjective majority vote of board-certified cardiologists interpreting the ECG with STEMI as the surrogate for OMI. 23 , 32–36 In addition, they often employed a spectrum of input clinical features in addition to the ECG waveform restricting their practical, real-world implementation.…”

International Evaluation of an Artificial Intelligence-Powered Ecg Model Detecting Occlusion Myocardial Infarction

“…4 Several previous studies deployed machine learning to triage patients presenting with ACS, however, bearing multiple limitations. 30,[38][39][40][41][42][43][44][45][46][47][48][49][50] The majority of these studies did not validate the occlusive or flow-limiting culprit lesions on coronary angiogram and relied on a subjective majority vote of board-certified cardiologists interpreting the ECG with STEMI as the surrogate for OMI. 30,[39][40][41][42][43] They often employed a spectrum of input clinical features in addition to the ECG waveform restricting their practical, real-world implementation.…”