Digitization of Electrocardiogram Using Bilateral Filtering

Electrocardiogram (ECG) is one of the most basic tools available for the detection of cardiovascular disease (CVD) which has evolved as one of the major causes of death in recent times, spanning countries. Conventionally, ECGs are printed on graph sheets which keep fading away with time. Hence, the preservation of ECG graph sheets is difficult. Digitization of printed ECG graph sheets is a solution to this problem. Digitization also helps in faster analysis and interpretation of ECG signals. Although the previous works have succeeded in smoothing ECG after extracting the signal from graph sheet, the cost amounts to loss of peak amplitude characteristics leading to wrong diagnosis. In this work, we propose a digitization technique to extract ECG signals from graph sheets. The proposed bilateral filter based method achieves a high degree of smoothness with all the peak characteristics intact when compared to other implementations using Butterworth filters. With this technique, it is now possible to create extensive databases similar to MIT-BIH from printed ECG graph sheets. A high level of automation is achieved, and results verified with over 60 ECGs and compared.