Medical Image Classification using Hybrid classifier by Extending the Attributes

Background/Objective: To create a Computer Aided Diagnosis system to detect the abnormalities in the human tissue images by extending the attributes. Methods/Statistical Analysis: An efficient and hybrid classifier using "K-Ratio Super Item set Finding-Nearest Neighborhood Classifier (KRSIF-NNC) Algorithm" is proposed. It classifies the tumor cells in an effective manner by adopting extended attributes from small datasets. The glioblastoma and lung cancer tissue image samples are keyed in to the algorithm which classifies them into four grades. Findings: From the histopathology (tissue) images the pathologists will be able to diagnose the abnormalities in the tissues. Examination and judgments are based on the pathologist's personal experience. The problem is during the manual diagnosis, there is a chance of missing some cancerous cells in the tissue images. This is solved by adopting the proposed classifier which automatically do the diagnostic process and classify it into proper grade. Thus the proposed classifier improves the classification process. Applications/Improvements: Improved classification is required to detect the cancer grades. This hybrid approach has better classification accuracy than other approaches with 4% improvement which is very essential.