A cholesterol lesion detection approach based on SVD decomposition

Fluorescence spectroscopy based on immuno-fluorescence is a useful tool in medical diagnosis and screening for different pathologies, as breast cancer, atherosclerosis ... Such a diagnosis, often achieved through a fluorescent labeling, is often obscured by a parasite signal associated mainly with some cells which are naturally auto-fluorescent. This paper proposes a computational method based on Singular Value decomposition which helps analyze the main components of an emitted fluorescent signal by a labeled artery. The aim of this analysis is to extract the pure spectra associated with the fluorescent sources and their relative repartition in order to detect and localize the cholesterol lesion. Computer simulations are presented to illustrate the efficiency of the proposed method in dealing with overlapped source problem in a labeled artery.