In recent years, the automatic identification and classification of tumor regions have gained more interest due to accuracy and reduced time complexity. One of the important strategies in tumor identification is segmenting the image as tumor and nontumor region, and this helps the researchers more significantly, as the MRI image comes in different modalities. This work introduces novel optimization based strategy for segmenting and classifying the image. Initially, the MRI images in the database are subjected to pre-processing and given to the segmentation process. For segmentation, this work utilizes the deformable model, and Fuzzy C Means (FCM) algorithm and the resultant segmented images are hybridized through proposed Dolphin based Sine Cosine Algorithm, preferred to be Dolphin-SCA. After segmentation, the tumor and non tumor-related features are extracted using the power LBP operator. The extracted features are subjected to Fuzzy Naive Bayes classifier for the classification, and finally, the classifier finds the suitable tumor class labels. Here, the entire experimentation is done by taking the MRI images from the BRATS database, and evaluated based on sensitivity, specificity, accuracy and ROC metrics. The simulation results reveal the dominance of proposed scheme over other comparative models, and the proposed scheme achieved 95.249% accuracy.