Automated Brain Tumor Detection Using Soft Computing-Based Segmentation Technique

Development and growth of abnormal cells within the brain results in brain tumor. In this study, a novel segmentation methodology is proposed for the segmentation of tumor. The proposed model consists of two phases. In the first phase, the brain CT image from the medical database is pre-processed to remove artifacts and noise. For Image segmentation, a Hierarchical Self Organizing Map (HSOM) is used that provides promising segmentation results. The conformist Self Organizing Map (SOM), which was used to categorize the picture row by row, is extended by the HSOM. Thus, the HSOM with vector quantization speeds up calculation at this lowest level of the weight vector, where there are more tumor pixels. The proposed automated system is tested on Kaggle (online available) database and achieves an accuracy of 98.94%.