Medical image mining techniques helps the medical world to identify diagnostic methods for prolonged and epidemic diseases. Peculiar disease like glioblastoma-type of brain tumor, challenges data/image mining technique with small datasets making it complex to classify. We propose an efficient and hybrid medical image classification method using "K-Ratio Super Itemset Finding-Nearest Neighborhood Classifier (KRSIF-NNC) Algorithm". The proposed mechanism expands the small dataset into subsets of different ratios pertaining to nucleus, cell and cytoplasmic structure of Glioblastoma Multiform (GBM) whole slide images. It then classifies them into different grades based on how serious glioma is-Grade I (benign-normal cells), Grade II (slightly abnormal cells), Grade III (malignant-cancerous cells) and Grade IV (most malignant cancerous cells). This algorithm can assist medical experts for effective classification and analysis of glioblastoma disease providing accurate diagnosis. Experimental results on single/multiple glioblastoma cell/cells image processing using the above algorithm proved efficient providing 98% accuracy and 97% sensitivity respectively. Data and intermediaries for a number of tumor types (GBM) are available at: http://tcga.lbl.gov for correlation with TCGA (The Cancer Genome Atlas) molecular data. © Research India Publications.