Signal Quality Assurance of ECG Signals for Automated Signal Processing Techniques

Assessing the ECG signal quality is necessary for the automated detection methods. If a noisy signal is sent to cardiologists or automated systems, it may lead to false positive or false negative interpretations which can lead to misdiagnosis. In this work, we have utilized heuristic rules of ECG signals and R-R interval based parameters to develop a novel algorithm to assess the signal quality. The signal quality assurance algorithm can be placed before the automated detection or classification algorithms to identify if the acquired signal is suitable for further processing or if it needs to be reacquired. The signal quality assurance algorithm has been validated using the MIT-BIH Normal Sinus Rhythm and MIT-BIH Noise Stress Test Databases. The algorithm has efficiently identified the signal below and equal to SNR of 12 dB as noisy signals and unacceptable for further signal processing. This work also identifies the impact of noise addition on feature point detection such as QRS, P and T waves which is imperative for any signal processing classification technique beginning from the traditional empirical methods to the most advanced neural networks approaches. Along with the database signals we have also validated it with ECG signals obtained in controlled laboratory settings by deliberately inducing noise by various means.