Energy-Aware Design Methodology for Myocardial Infarction Detection on Low-Power Wearable Devices
Energy-Aware Design Methodology for Myocardial Infarction Detection on Low-Power Wearable Devices
复制标题
DOI：
10.1145/3394885.3431513
复制
发表时间：
2021-01
期刊：
影响因子：
--
通讯作者：
Mohanad Odema;Nafiul Rashid;M. A. Faruque
中科院分区：
文献类型：
--
作者：
Mohanad Odema;Nafiul Rashid;M. A. Faruque
心肌梗塞 (MI) 是一种损害心肌的心脏病，需要立即治疗。它的无声和反复发生的性质需要对患者进行实时连续监测。如今，可穿戴设备足够智能，可以在设备上对心跳片段进行处理并报告其中的任何异常情况。然而，可穿戴设备的小外形尺寸带来了资源限制，并且需要节能解决方案来满足这些限制。在本文中，我们提出了一种设计方法来自动探索用于 MI 检测的神经网络架构的设计空间。该方法结合了神经架构搜索 (NAS)，使用多目标贝叶斯优化 (MOBO) 来呈现帕累托最优架构模型。这些模型最大限度地减少了目标设备上的检测误差和能耗。该设计空间的灵感来自适用于资源有限的移动健康应用的二元卷积神经网络（BCNN）。使用 PhysioNet 的 PTB 诊断心电图数据库验证模型的性能。此外，能源相关的测量结果是通过典型的硬件在环方式直接从目标设备获得的。最后，我们将我们的模型与其他相关工作进行基准测试。一种模型超过了可穿戴设备上最先进的精度（达到 91.22%），而其他模型则牺牲了一些精度来降低能耗（达到 8.26 倍）。
Myocardial Infarction (MI) is a heart disease that damages the heart muscle and requires immediate treatment. Its silent and recurrent nature necessitates real-time continuous monitoring of patients. Nowadays, wearable devices are smart enough to perform on-device processing of heartbeat segments and report any irregularities in them. However, the small form factor of wearable devices imposes resource constraints and requires energy-efficient solutions to satisfy them. In this paper, we propose a design methodology to automate the design space exploration of neural network architectures for MI detection. This methodology incorporates Neural Architecture Search (NAS) using Multi-Objective Bayesian Optimization (MOBO) to render Pareto optimal architectural models. These models minimize both detection error and energy consumption on the target device. The design space is inspired by Binary Convolutional Neural Networks (BCNNs) suited for mobile health applications with limited resources. The models’ performance is validated using the PTB diagnostic ECG database from PhysioNet. Moreover, energy-related measurements are directly obtained from the target device in a typical hardware-in-the-loop fashion. Finally, we benchmark our models against other related works. One model exceeds state-of-the-art accuracy on wearable devices (reaching 91.22%), whereas others trade off some accuracy to reduce their energy consumption (by a factor reaching 8.26×).