Snippet Policy Network for Multi-class Varied-length ECG Early Classification

Arrhythmia detection from ECG is an important research subject in the prevention and diagnosis of cardiovascular diseases. The prevailing studies formulate arrhythmia as a time series classification problem. Meanwhile, early presents real-world demand for diagnosis. In this paper, we address problem diseases classification, which varied-length long-length well. For solving problem, propose deep reinforcement learning-based framework, namely Snippet Policy Network (SPN), consisting four modules, snippet generator, backbone network, controlling agent, discriminator. Comparing to existing approaches, proposed framework features flexible input length, solves dual-optimization solution earliness accuracy goals. Experimental results demonstrate that SPN achieves excellent performance over 80% terms accuracy. Compared state-of-the-art methods, at least 7% improvement on different metrics, including precision, recall, F1-score, harmonic mean, delivered by SPN. To best our knowledge, first work focusing based data. Based these SPN, it offers good exemplification addressing all kinds problems.