

In the clinical ophthalmology, analysis of the retinal images is a routine process having a goal the proper clinical diagnosis assessing the retinal state. Unfortunately, physicians often perform the diagnosis subjectively based on their opinion which is strongly depended on their experience, and skills. Furthermore, such approach does not allow the diagnostic parameter accurate quantification. In this regard, the automatic modeling allowing for the clinical retinal features extraction is substantially important. We have proposed a fully automatic segmentation procedure based on the active contour driven by the Gaussian distribution with the goal of the retinal lesions modeling. Since the retinal lesions have approximately stable brightness spectrum without significant variations, the Gaussian energy approximation appears as an effective approach. Active contour consecutively adopts a shape of the retinal lesions whilst its energy is being minimized. Eventually, we obtain a closed and smoothed curve reflecting manifestation and geometrical features of the individual retinal lesions. On the base of the segmentation model, we have selected two clinically significant features which are tracked over the time via the linear regression model. This model allows for tracking the dynamical variations of the retinal lesions which are not stable, as it is supposed. This system allows for automatic modeling, and prediction system for the retinal lesions image data.