A novel feature extraction approach for tumour detection and classification of data based on hybrid SP classifier

Abstract

This paper deals with identifying the cancer affected region of the brain. Many tools and techniques such as self-organising map (SOM), Proximal Support Vector Machine (PSVM) classifiers etc. exist to find out the cancer affected region in the brain. But the rapid growth in brain tumour cases in recent past indicates that the existing technologies have failed to identify its root cause as identification is a complex process and recent studies also reveal that different types of brain tumours can be treated either through surgery or in rare cases, with radiation. Image segmentation helps in identifying brain tumours, by calculating the volume and the growth of the tumours using techniques like human edge correction, outer edge colouring and interactive threshold holdings. In order to reduce the human error and to get the accurate results in MRI images there is an urgent need to find out an automatic or semi-automatic method for the classification of brain tumour images. The paper presents a 'hybrid SP' classifier and discusses its results in the detection and classification of brain cancer.