An efficient approach for detecting brain tumors using a modified artificial neural network

Abstract

ABSTRACT Computed Tomography (CT) was a prominent technique for detecting brain tumours. Tradition models have some drawbacks, such as that they do not predict the accurate position of the tumour, are time-consuming, involve complex operations, and have low accuracy. In order to solve these issues, Modified Artificial Neural Network (ANN) was proposed to detect brain tumour in a CT image. Initially the raw images were pre-processed to improve contrast and reduce the noise using contrast stretching and median filtering. Then the features were extracted from the pre-data by using gradient filter, Local Binary Pattern, Gabor feature extraction, and Histogram of oriented gradients. The extracted features were given to the modified ANN model to predict the disease. The prediction performance of the classifier was improved by selecting the appropriate weight of the ANN. The optimal selection of weight was done through the use of sailfish optimisation algorithm. The proposed method obtained 94% accuracy, while existing systems such as SVM, KNN, RF, and NB reached 90%, 85%, 74% and 72% accuracy, respectively, which is lower than the proposed method. The proposed early and accurate diagnosis of brain tumour model thus effectively detects the tumour and provides necessary treatment based on this diagnosis.