Key result
Optimized neural network achieves ~99.6% arrhythmia detection accuracy, matching traditional models with reduced computational complexity.
Why the study?
Due to lack of human expertise and high probability of misdiagnosis in arrhythmia detection from ECG signals, computer-aided diagnosis methods are preferred.
An optimized neural network and self-learning method can achieve high accuracy (up to 99.56%) in computer-aided arrhythmia detection from ECG signals.
May facilitate efficient ECG arrhythmia detection; leaves open prospective clinical validation before adoption.
Electrocardiogram (ECG) is one of the non-invasive and low-risk methods to monitor the condition of the human heart. Any abnormal pattern(s) in the ECG signal is an indicative measure of malfunctioning of the heart, termed as arrhythmia. Due to the lack of human expertise and high probability to misdiagnose, computer-aided diagnosis and analysis are preferred. In this work, we perform arrhythmia detection with an optimized neural network having piecewise linear approximation based activation function to alleviate the complex computations in the traditional activation functions. Further, we propose a self-learning method for arrhythmia detection by learning and analyzing the characteristics (period) of the ECG signal. Self-learning based approach achieves 97.28% of arrhythmia detection accuracy, and neural network with optimized activation functions achieve an arrhythmia detection accuracy of 99.56%.
No takes yet. Share an insight, caveat, or question.
Dinakarrao et al. (2018) studied Arrhythmia (n=48). Optimized neural network with piecewise linear approximated activation functions (PLAtanh) vs. Traditional neural network with exact hyperbolic tangent functions (tanh) was evaluated on Arrhythmia detection accuracy. An optimized neural network with piecewise linear approximated activation functions achieved an arrhythmia detection accuracy of 99.56%, comparable to traditional neural networks but with reduced computational complexity.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: