Premature Ventricular Complex Detection Chip Obtained Using Convolution Neural Network
Yu‐Lun Huang, Pei-Jung Chang, Yuan‐Ho Chen
Abstract
In this work, a very large scale integration chip for arrhythmia detection is proposed. The proposed chip comprises a convolution neural network (CNN) for detecting the abnormal heartbeat of premature ventricular complex (PVC) by using 4-lead electrocardiogram signals. The proposed CNN comprised two convolution layers and one fully connected layer to achieve high-accuracy PVC detection. The detection accuracy of the proposed CNN circuit was 94.94% when implemented in a single chip using TSMC <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$0.18-\mu\ \mathbf{m}$</tex> complementary metal-oxide-semiconductor processing techniques. The results indicated that the proposed core consumes 3.1-mW power with a clock frequency of 66.6 MHz, and the gate count of the proposed core was 14 K. Thus, the proposed CNN offers high speed, a small area, and high-accuracy PVC detection.