We introduce a new methodology for linear unmixing of optic signals, which is denoted as extended blind end-member and abundance extraction. To validate the proposal, we study m-FLIM datasets to classify fresh postmortem human coronary atherosclerotic plaques.