Certain investigations on efficient segmentation and classification approach for identifying the brain tumour from mr imaging

Abstract

The brain tumour is the leading cause of cancer-related death worldwide and the high death rate is due to advanced stages of cancer at diagnosis. The development of automated classification helps the physicians to diagnose the brain tumour accurately based on the symptoms and MRI scan images in its early stage. Various techniques are available for brain tumour detection, but many methods provide less accuracy and more false positives. Therefore, there is a need to improve the tumour detection mechanism and to deliver high-level treatment to current therapies. The study focuses on understanding the brain tumour classification mechanism and implements a novel method to detect brain tumour. The main objectives of this research are to develop a classification and segmentation framework for brain tumour detection from the given MRI scan images. In the present work, a system for the classification of brain tumour for early detection of cancer using watershed segmentation and threshold techniques were developed. The watershed segmentation is employed along with the aid of grayscale image on the MRI image trailing the threshold and morphological operator for identifying tumours. This will separate the features in terms of shape, size and segments of the brain. Further, an improved version for segmentation is proposed with the Adaptive Histogram Balancing analysis and Gray Level Co-Occurrence Matrix (GLCM) is deployed for feature extraction. After that, the segmentation is carried out by using the framework called EnhancedProbabilistic Neural Network. The accuracy of the proposed method is verified using a confusion matrix. Finally, the Improved Classification Model for Brain Tumour Diagnosis (ICM-BTD) method is proposed. Initially, the acquired scan images are used for filtering and applied over a feature extraction process done by Gray Level Co-Occurrence Matrix (GLCM) and Discrete Wavelet Transform. At last, the classification is done with the technique called Support Vector Machine (SVM). Since the accuracy is improved greatly with minimal processing time. The implementation part is completely verified by using MATLAB tool. The dataset is downloaded from DICOM library website for experimental purpose. The collective MRI scan images with normal and abnormal images were considered for processing. Finally, the observations are stated that the proposed ICM-BTD is better than the other methodologies like FCM and PNN. The precision values are represented as 96% and accuracy witnessed as 94% on average newline