What is the clinical evidence from this study?

Study design: Other. Population: Arrhythmia (n=3). Intervention: TinyCES (TinyML-based classification) vs. Real-time feedback schemes (RTFs) and conventional Holter. Primary outcome: Classification accuracy (detection ratio).

January 1, 2023Computers, materials & continua/Computers, materials & continua (Print)86 citationsOpen Access

TinyML-Based Classification in an ECG Monitoring Embedded System

EKEunchan KimHanyang University JKJaehyuk KimHanshin University JPJuyoung ParkMokpo National University

Key Result

The TinyCES embedded ECG monitoring system achieved approximately 97% classification accuracy while reducing network bandwidth usage by 96.71% and memory usage by 96.23% compared to conventional methods.

Structured PICO

Population

Three healthy volunteers evaluated a TinyML-based ECG monitoring prototype for 24 hours to assess real-time classification and resource usage.

Intervention

TinyCES: a tiny machine learning (TinyML)-based classification system using a CNN model on an Arduino Nano 33 BLE Sense microcontroller for real-time on-device ECG analysis.

Comparator

Real-time feedback schemes (RTFs) transmitting data via BLE to a smartphone, and conventional Holter monitoring.

Outcome

Classification accuracy, average amount of network bandwidth usage, and average size of recorded ECG data.surrogate

Implementing TinyML directly on an ECG monitoring embedded device achieves 97% arrhythmia detection accuracy while drastically reducing network bandwidth and memory requirements.

Limitations

Accuracy depending on specific types of abnormal ECG segments (e.g., atrial fibrillation) needs further validation
Requires further experimental validation with heart disease patients in a hospital setting

Abstract

Recently, the development of the Internet of Things (IoT) has enabled continuous and personal electrocardiogram (ECG) monitoring. In the ECG monitoring system, classification plays an important role because it can select useful data (i.e., reduce the size of the dataset) and identify abnormal data that can be used to detect the clinical diagnosis and guide further treatment. Since the classification requires computing capability, the ECG data are usually delivered to the gateway or the server where the classification is performed based on its computing resource. However, real-time ECG data transmission continuously consumes battery and network resources, which are expensive and limited. To mitigate this problem, this paper proposes a tiny machine learning (TinyML)-based classification (i.e., TinyCES), where the ECG monitoring device performs the classification by itself based on the machine-learning model, which can reduce the memory and the network resource usages for the classification. To demonstrate the feasibility, after we configure the convolutional neural networks (CNN)-based model using ECG data from the Massachusetts Institute of Technology (MIT)-Beth Israel Hospital (BIH) arrhythmia and the Physikalisch Technische Bundesanstalt (PTB) diagnostic ECG databases, TinyCES is validated using the TinyML-supported Arduino prototype. The performance results show that TinyCES can have an approximately 97% detection ratio, which means that it has great potential to be a lightweight and resource-efficient ECG monitoring system.

Ask AI

Helpful

Bookmark

View Full Paper