Soft Computing based Brain Tumor Categorization with Machine Learning Techniques

Abstract

When it comes to medical science, radiology is a wide area that demands further information and thought in order to execute an appropriate tumour examination. This study makes use of MRI sequence pictures as input images to identify the tumour site, and as a consequence of this work, a malignant segment and detection technique is established. This expression is difficult to perform because of the considerable variability in the presence of cancer tissues linked with different inmates, as well as the similarity among normal tissues in the majority of cases, which makes the task difficult to finish. The most significant goal is to divide the brain into two groups: those with malignant tumours and those who do not have malignant tumours. There are four primary phases in the system that is presented. For efficient malignant detection, the registration process is carried out first using Edge based Contourlet Transformation, followed by segmentation of tumour points using region-expanding segmentation, followed by aspect extraction using two types of texture features, namely Otsu Thresholding, K-means, and Local Binary markings texture aspect, and finally, classification using neural network methods is imported out. Using reverse propagation detection of malignancy from a Slices scan image, the proposed approach is a one-of-a-kind procedure that may be used to identify the existence of tumours. For classification, a backpropagation method was utilised, and the accuracy of the classification was increased as a result. A variety of MRI sequences are used to test the proposed technique, which is implemented in Mat lab and yields experimental results for Image Registration and segmentation using point of growth. When the segmented photographs are compared to the victims' database, a method called Backpropagation neural network classification is used to classify them as serious or benign, respectively.