Computationally Efficient LOB Algorithm for Digitization of Degraded Paper ECG Reports

Abstract

Purpose: A less computationally intensive methodology is required to digitise paper Electrocardiogram (ECG) records from scanned photographs so that it can be implemented on mobile devices and used in remote places without the necessary hardware or software requirements. Due to numerous degradations inherent in older resting paper ECG recordings, automated digitization of paper ECG records also becomes challenging. Methods: In the proposed method, we present a robust and simple, Level of Binarization (LOB) characteristic curve to binarize paper records. The binarized image is processed using the discontinuous signal correction (DSC) technique and is vertically scanned to extract 1-D ECG signal. Results: The proposed method is tested for paper ECG records having degradations like variable background colour, steep slope, ink evaporation, and folding. The average RMS error and accuracy for 101 test samples, is 0.01 and 98.34 % respectively. The results show that, the LOB-DSC gives better performance when compared to the existing ECG extraction methods. As proposed, the LOB method does not require pre-processing, and the computational complexity is also reduced. Conclusion: An automatic threshold for binarization and estimation of several types of degradations in old ECG records is obtained using the less computationally intensive LOB characteristic curve. This method can be applied in remote locations with low availability of advanced processing devices for ECG signal extraction.