Synthesis of ECG from arterial blood pressure and central venous pressure signals using Artificial Neural Network

In this context, the synthesis of ECG cycles from arterial blood pressure (ABP) and central venous pressure (CVP) signals using Artificial Neural Network (ANN) is described. The proposed method utilizes synchronously sampled ABP and CVP cycles of a patient for the generation of ECG cycles of that patient. The signals in the study are taken from MGH/MF waveform database. The radial basis neural network is trained by segmenting the input and target signals into smaller segments of equal length consisting of 2500 samples. This trained ANN outputs ECG lead-II signals with independent ABP and CVP signals as input. The generated ECG signals possess resemblance with actual ECG signals available from the database. The accuracy of this generated ECG is given in terms of cosine measure and cross correlation coefficient with respect to actual ECG.