Abstract : T waves properties are not usually fully exploited when analyzing its profile variation within a set of records. An affine transform of the temporal support is a good candidate for the modeling of this variation. Based on this assumption, we propose here a new framework that provides an estimation of the reference T wave and the parameters of the affine functions. To achieve this task, the domain of the inverse normalized integral is investigated, allowing a compact description of the dataset. As a straightforward tool, the Principal Component Analysis is shown to be efficient in this do- main. This approach is illustrated by analyzing a set of T waves recorded during Dipyridamol infusion and during angioplasty.