T waves segmentation and analysis using inverse normalized integrals

Stress tests are typical protocols that exhibit T waves changes for healthy and ischemic subjects. ST-T elevation is one example among others that characterizes ischemic patient. A more complex description of T wave changes needs extended models with a cost of higher identification pitfalls. We propose here to work in the inverse normalized integrals domain. This domain permits us an estimation of all pertinent parameters for the characterization of T waves shape. These parameters stand for the delay, the width (defined in the paper as the scaling factor) and the offset. In contrast, direct application of PCA on the data (defined in the paper as time-domain approach) assumes perfectly aligned waves with identical widths. Time-varying parameters values corresponding to exercise tests are estimated by using both approaches. It is shown that results are consistent with clinical knowledge.