Efficient Detection of Brain Tumor Using Unsupervised Modified Deep Belief Network in Big Data

Ramdas Vankdothu, Dr. Mohd Abdul Hameed and Husnah Fatima

The presence of abnormal cells in the brain induces a Brain tumor. Brain tumors are the result of abnormal increases and independent cell movement in the brain. There are two types of tumors in the brain those classified as the slow-growing tumor, also called Low grade and fast-growing tumors, also known as High-grade. The patient's efficient operation depends on tumor discovery Accuracy. False detection accuracy can lead to the death of the patient. It is helpful to patient life if they detected early and correctly. Some types of brain tumors, such as Meningioma, Glioma, and Pituitary tumors, are more popular than the others. Nowadays, deep learning has been performing a significant part in the area of computer vision. Recently One of the more critical tasks on the brain tumor detection method recommended on an adaptive convolutional neural network model CNN-BN-PReLU based on the convolutional neural network approach. It first executes the batch normalization (BN) Process on the input of every feature map of each layer of the network. However, the obtained results are unstable, lack high accuracy, and the adaptability is not reliable. To end this, in this paper, we are presenting a solution for the detection of Brain tumor segmentation using the Modified Deep Belief Network (MDBN) method from MRI images. Mainly we are focusing on feature extraction to provide satisfactory completion in our feature extraction for all the images. The suggesting feature extraction method is, based on a statistical test preceded by a Principal component analysis (PCA). Finally, a comparative study presents between, support vector machine (SVM), artificial neural network (ANN), and Modified Deep Belief Network (MDBN) based unsupervised algorithms, results are compared to different criteria for sensitivity, specificity, and accuracy of all works. The proposed method provides better performance with 97.53% sensitivity, 95.82% privacy and 99% accuracy across the entire Brain Atlas database.

Volume 12 | 04-Special Issue

Pages: 338-347

DOI: 10.5373/JARDCS/V12SP4/20201497