Exploiting the underlying cepstral coefficients for large scale and fine-tuned EKG time-imagery analysis including R-R, P-R, R-T, and R-PVC interval imaging

This paper presents a three-part expansion for feature extraction of the electrocardiogram (EKG) based on a deeper exploitation of what cepstral processing can reveal, specifically in two distinct forms of time-varying-cepstral-power images. These two images have been reviewed for readability by qualified EKG technicians and have received positive reviews in light of their perception of having a tremendous gain in large-scale, yet fine-tuned, EKG strip interpretation ability. Our demonstration gives a brief review of EKGs in the time domain before moving to the image domain where we emphasize the pattern recognition aspects of our approach. Section II of this paper defines, and gives real biological applications for, the cepstrum and introduces our framing style for visualization. We give an example of a voice signal from one of the authors in this paper. Section III of this paper demonstrates our framing technique using two EKG records. Here, we lay the groundwork for imaging EKG as an image of EKG cepstral-feature terrain in the context of R-R, P-R, and R-T, as well as R-PVC intervals since this record contains premature ventricular contractions (PVCs) in addition to normal sinus rhythm (NSR). The first record (named record 105) was obtained from the Physionet MIT-BIH (MITDB); the second record was obtained from the Clemson University Biomedical Lab and is the EKG of another author of this paper. Section IV of this paper applies what we have learned from the first two parts to a variation of recurrence plots. In this instance, the recurrence plot itself is imbedded within a correlation matrix formed from a cepstral domain image. In this way, we point to our next stages of pattern recognition that involve yet another layer of exploitation of (dynamic) sub-matrices through 2-D segmentation and extraction which, when placed back in context of time domain, may further reveal clues to understanding the complex and seemingly acyclical nature of heart rate reset.