Artificial intelligence-assisted remote detection of ST-elevation myocardial infarction using a mini-12-lead electrocardiogram device in prehospital ambulance care

Abstract

Objective: To implement an all-day online artificial intelligence (AI)-assisted detection of ST-elevation myocardial infarction (STEMI) by prehospital 12-lead electrocardiograms (ECGs) to facilitate patient triage for timely reperfusion therapy.

Methods: The proposed AI model combines a convolutional neural network and long short-term memory (CNN-LSTM) to predict STEMI on prehospital 12-lead ECGs obtained from mini-12-lead ECG devices equipped in ambulance vehicles in Central Taiwan. Emergency medical technicians (EMTs) from the 14 AI-implemented fire stations performed the on-site 12-lead ECG examinations using the mini portable device. The 12-lead ECG signals were transmitted to the AI center of China Medical University Hospital to classify the recordings as "STEMI" or "Not STEMI". In 11 non-AI fire stations, the ECG data were transmitted to a secure network and read by available on-line emergency physicians. The response time was defined as the time interval between the ECG transmission and ECG interpretation feedback.

Results: Between July 17, 2021, and March 26, 2022, the AI model classified 362 prehospital 12-lead ECGs obtained from 275 consecutive patients who had called the 119 dispatch centers of fire stations in Central Taiwan for symptoms of chest pain or shortness of breath. The AI's response time to the EMTs in ambulance vehicles was 37.2 ± 11.3 s, which was shorter than the online physicians' response time from 11 other fire stations with no AI implementation (113.2 ± 369.4 s, P < 0.001) after analyzing another set of 335 prehospital 12-lead ECGs. The evaluation metrics including accuracy, precision, specificity, recall, area under the receiver operating characteristic curve, and F1 score to assess the overall AI performance in the remote detection of STEMI were 0.992, 0.889, 0.994, 0.941, 0.997, and 0.914, respectively. During the study period, the AI model promptly identified 10 STEMI patients who underwent primary percutaneous coronary intervention (PPCI) with a median contact-to-door time of 18.5 (IQR: 16-20.8) minutes.

Conclusion: Implementation of an all-day real-time AI-assisted remote detection of STEMI on prehospital 12-lead ECGs in the field is feasible with a high diagnostic accuracy rate. This approach may help minimize preventable delays in contact-to-treatment times for STEMI patients who require PPCI.

Keywords: ST-elevation myocardial infarction (STEMI); artificial intelligence (AI); contact-to-balloon (C2B) time; convolutional neural network and long short-term memory (CNN-LSTM); prehospital 12-lead ECGs.

Figure 1

The flowchart of the AI-based pre-hospital STEMI detection system. Traditionally, after the 12-lead ECG had been recorded in the ambulance vehicle, the ECG data were posted on a secured network for reading by available online physicians as had been usual practice. The time interval between ECG transmission and interpretation feedback by physicians was defined as the physician's response time. In our AI-based pre-hospital STEMI detection system, the recorded signal was also simultaneously transmitted to the AI center of the China Medical University Hospital to be classified “STEMI” or “Not STEMI.” Similarly, the time interval between the ECG transmission and the ECG interpretation feedback by the AI was defined as the AI's response time.

Figure 2

Diagram depicting the deep learning model architecture. The deep learning model used a combination of CNN and LSTM to classify STEMI on 12-lead ECGs. The architecture of the proposed AI mode was composed of two 1D -CNNs blocks fed with chest and limb leads, to extract the features from the 6-lead signals. The outputs of the two 1D-CNNs were connected to two layers of LSTM, which served as a sequence analyzer. Then the outputs of the two LSTMs were concatenated and connected to a fully connected layer to classify the data as “STEMI” or “Not STEMI”.

Figure 3

Validation of 12-lead ECG signals between devices. The proposed deep learning model for STEMI detection was based on the digital 12-lead ECG signals recorded using a computerized ECG machine (GE Healthcare MAC 2000/3500/5500, USA). For the prehospital 12-lead ECG acquisition in the current study, we used a mini portable ECG device (QT Medical, Diamond Bar, CA, USA) with the proposed AI algorithm integrated within. To ensure the efficacy of AI-based STEMI detection using the mini-portable ECG device, we checked the consistency of the 12-lead ECG output signals between the two devices. A total of 194 verified ECGs acquired from GE machines (GE-ECGs) were converted into the corresponding QT Medical ECG output format (QT-ECGs) using a certified ECG simulator. Finally, the signal similarities between raw GE-ECGs and the transcribed QT-ECGs was analyzed. The performance of the AI model in classifying data as “STEMI” or “Not STEMI” for each of the two sets of ECG signals was compared to attest to the consistency of AI performance across devices.

Figure 4

Comparison of response time between AI and physicians on prehospital ECGs. Between July 17, 2021, and March 26, 2022, the proposed AI model classified a total of 362 prehospital 12-lead ECGs as “STEMI” or “Not STEMI”, obtained from 275 consecutive patients who had called the 119 dispatch centers of the fire stations in Taichung City and Nantou County for symptoms of chest pain or shortness of breath. The AI's response time to the EMTs in ambulance vehicles was 37.2 ± 11.3 s, which was shorter than the online physicians' response time (113.2 ± 369.4 s, P < 0.001) from 11 other fire stations with no AI implementation after analyzing another set of 335 prehospital 12-lead ECGs.

Figure 5

Representative ECGs of false positive and false negative labeling by the proposed AI model. (A) This prehospital ECG showing pathological Q waves in the inferior and anterolateral leads was classified as STEMI by AI. The ground truth committee interpreted this ECG as recent or old myocardial infarctions and judged this AI labeling as a false positive case. (B) There was only one ECG with a false negative labeling by AI as “Not STEMI,” which was interpreted as “STEMI” according to the adjudication by the ground truth committee. Interestingly, this patient was diagnosed with a recent myocardial infarction by the cardiologist in charge at the destination hospital after incorporating more hospital-based information including historical ECGs and laboratory data, and did not require PPCI. (C) The evaluation metrics including area under the receiver operating characteristic curve, accuracy, specificity, precision, recall, and F1 score to assess the overall AI performance in the remote detection of STEMI from 362 prehospital 12-lead ECGs were 0.997, 0.992, 0.994, 0.889, 0.941, and 0.914, respectively.