The proposed method predicts the grade of astrocytoma based on certain features extracted from magnetic resonant images. MR brain images of three views namely Axial, Sagittal and Coronal are considered for perfect grade evaluation of astrocytoma. Gray level co-occurrence matrix is utilized for feature extraction. The extracted features are subjected to classification through a classifier which predicts the grade. Fuzzy logic classifier and K-Clustering network are the most used classifiers for prediction of grades among many available. The performances of the techniques are evaluated through discussions with neuroradiologist to find their accuracies. A modified K-Clustering network is developed to give a better diagnosis for predicting the grades of astrocytoma. © Springer India 2015.