The real-time monitoring of different molecular interactions can be used as a lower cost tool for genetic diagnosis. The extraction of the hybridization signal allows the estimation of the association/dissociation constants, the affinity of biomolecular components (target/probe) that interact and then characterize their activities and functions. This extraction of the biological information is based on the analysis of images acquired by a CCD camera during the course of the experiment and a self-calibration of the data obtained. Until now, the processing of these images was post experimental and concerned different stages of analysis: the detection of spots region, spatiotemporal segmentation of areas of interaction and eventually the quantification of these areas using the kinetic response measured. The challenging issue is to continue to improve the automatic extraction of the interaction signal and develop a processing tool applied in real-time as the image acquisition progresses. The advantage of such treatment is to allow the prediction of the evolution of the interaction, especially in the case of genetic diagnosis. It may also detect any malfunction that may arise during the interaction and allow the experimenter to decide whether to continue or interrupt the experience. This paper proposes a new approach for the real-time analysis of the image data provided by the SPR. A self-calibration step allows the correction of microarray design flaws or of temporal artifacts. Once the data are normalized, 3D morphological operators are used to extract the binary mask that will allow detecting all regions of interest for dynamic segmentation. This segmentation is then used in a spatio-temporal classification to extract the effective signal within each detected spot. The resulting real-time analysis approach presents a great interest in genetic diagnosis applications.