Abstract 16611: The Feasibility of Artificial Intelligence in Interpretation of the Treadmill Test

Introduction: The treadmill test (TMT) is commonly used as a screening test to identify myocardial ischemia. The sensitivity of the TMT in diagnosing obstructive CAD was approximately 50%, with a specificity of approximately 90%. However, the challenge with TMT is the variability in interpretation.

Hypothesis: We aimed to develop an Artificial intelligence (AI) to interpret TMT. The primary endpoint was the accuracy comparison between the initial assessment (of AI or cardiologist) for obstructive CAD.

Methods: Out of a total of 800 TMTs performed on adults for any clinical indications in 2018-2020, Of these, we used 500 as training data, 200 as validation data, and 100 as test data. A one-dimensional convolutional block was applied to each of the 12 leads of the TMT, and then all the leads were concatenated to apply a dense layer, which was then trained to predict the presence of obstructive coronary artery disease. This deep learning model was developed based on convolutional neural networks (CNNs) using TensorFlow (Google, Mountain View, CA, USA) and Keras packages based on the Python language. TMT were simultaneously interpreted by an AI algorithm or by the cardiologist. This assessment was subsequently reviewed by blinded interventional cardiologist who provided a final report of obstructive coronary artery disease (CAD) with coronary compute tomography.

Results: The mean age was 58.1 (20-78) years, 68% were male, and 33% had obstructive coronary artery disease (≥70% stenosis in at least one vessel). The performance of AI was compared with that of conventional TMT. Accuracy of AI model was comparable to cardiologist (0.720 vs. 0.680, p =0.845) F1 score was also similar between AI and cardiologist (0.440 vs. 0.384. p =0.845)

Conclusions: Artificial intelligence is comparable to a cardiologist in the initial interpretation of the treadmill test. These approaches might be applicable to data obtained from wearable electrocardiography technologies during exercise.