An automated detection and segmentation of tumor in brain MRI using artificial intelligence

Abstract

Medical image segmentation is a crucial process which makes possible, the characterization and visualization of the structure of interest in medical images. Brain MRI segmentation is a more difficult procedure because of inconsistency of abnormal tissues like tumor. In this paper, we propose a fully automated technique that uses artificial intelligence to detect and segment abnormal tissues like tumor and atrophy in brain MRI images accurately. Three stages are offered in our work: (1) Feature Extraction (2) Classification and (3) Segmentation. The extracted features like energy, entropy, homogeneity, contrast and correlation from the brain MRI images are applied as input to an artificial intelligence system that uses a Neuro-fuzzy classifier which classifies the images into normal or abnormal. The abnormal tissues like tumor and atrophy are then segmented using region growing method. The accuracy of the segmentation results are assessed with metrics like False Positive Ratio (FPR), False Negative Ratio (FNR), Specificity, Sensitivity and Accuracy. This entire procedure is developed as a Graphical User Interface (GUI) system which results in automated detection and segmentation of tumor.