Template Matching Based Early Exit CNN for Energy-efficient Myocardial Infarction Detection on Low-power Wearable Devices

Myocardial Infarction (MI), also known as heart attack, is a life-threatening form of disease that leading cause death worldwide. Its recurrent and silent nature emphasizes the need for continuous monitoring through wearable devices. The device solutions should provide adequate performance while being resource-constrained in terms power memory. This paper proposes an MI detection methodology using Convolutional Neural Network (CNN) outperforms state-of-the-art works on devices two datasets - PTB PTB-XL, energy memory-efficient. Moreover, we propose novel Template Matching based Early Exit (TMEX) CNN architecture further increases efficiency compared to baseline maintaining similar performance. Our TMEX achieve 99.33% 99.24% accuracy dataset, whereas PTB-XL dataset they 84.36% 84.24% accuracy, respectively. Both architectures are suitable requiring only 20 KB RAM. Evaluation real hardware shows our 0.6x 53x more energy-efficient than improves by 8.12% (PTB) 6.36% (PTB-XL) architecture.