The optimal transport theory enables great flexibility in modeling problems associated with image registration, as different optimization resources successfully used because the choice of suitable matching models to align the pictures. The proposed method in this paper is an automated framework for multimodal fundus image registration using both colored and gray scaled images and graph matching schemes into a functional and easy method. Then, this method is used to predict the diseases accurately. Then this method is also used to predict whether the disease is affected or not affected by using a comparative method. These methodologies are validated by a comprehensive set of comparisons against competing and well-established image registration methods, by using real medical datasets and classic measures typically employed as a benchmark by the medical imaging community our proposed method is mostly used in medical field. It is used to easily detect the diseases. We demonstrate the accuracy and effectiveness of this framework throughout a comprehensive set of qualitative and quantitative comparisons against several influential state-of -the- art methods on various fundus image databases.