Regularized Anisotropic Filtered Tanimoto Indexive Deep Multilayer Perceptive Neural Network learning for effective image classification

Abstract

Image classification is a significant way in the field of image processing to automatically categorize large numbers of images. Brain tumor classification is mainly a helpful and widely desired process in the medical system. Brain tumor classification is a significant way to automatically categorize brain tumors images. Many methods have been introduced for solving the classification task with leverage low-level features. However, it has few limitations for achieving the higher accuracy of image tumor classification with minimum time. To overcome the issue, this study has proposed a novel technique called Regularized Anisotropic Filtered Tanimoto Indexive Deep Multilayer Perceptive Connectionist Network (RAFTIDMPCN), consisting of many layers of nodes for deeply analyzing the input and providing better classification results. The proposed architecture helps in improving the accuracy and reducing the time. The input layer receives the number of MRI images and natural image datasets collected from the dataset. Then the images are sent to the first hidden layer where the preprocessing is carried out to improve the image quality by removing the noise pixels using Regularized Anisotropic diffusion filtering technique. Followed by, shape, color, texture, and size features of input images are extracted in the second hidden layer. The classification is performed at the third hidden layer based on the Tanimoto similarity measure. Finally, the Heaviside step activation function is applied to obtain the classification results with higher accuracy. Experimental evaluation is carried out with different qualitative and quantitative results discussion by using brain tumor MRI dataset and natural image datasets. The obtained results indicate that the proposed technique provides better results in terms of Peak signal to noise ratio, accuracy, false-positive rate, time complexity, and space complexity. The analyzed results show the superior performance of our proposed RAFTIDMPCN technique accuracy by 6%, minimizes the false-positive rate by 40%, and time complexity by 11% in brain tumor detection when compared with the two state-of-the-art methods. This paper also presents several discoveries that could be helpful to the neurological community.