Medical image analysis is the most emerging and challenging field nowadays, among which brain tumor detection and classification is the most horrible and devastating type of cancer. Brain tumor classification helps in diagnosing the tumor at an early stage. Though numerous tumor detection methods and classification have been proposed, enhanced tumor detection and classification is still a challenging task because brain tumor images possess high diversity in boundaries as well as tumor appearance. This paper proposes a quick method for extracting metabolite values from graphs. Brain tumors are detected using metabolites such as N-acetylaspartate (NAA), choline, and Creatine ( ). The tumor type is decided by taking Choline /N-acetylaspartate ratio. During the clustering process, each metabolite is assigned a weight. This clustering strategy has a precision of up to 88 percent. The suggested method is built on decision tree algorithms, which outperform clustering algorithms. Instead of storing fMRI (functional MRI) mages the proposed theory stores values of metabolites in a dataset which improves the image processing tasks as well as memory requirements.