An Improved Deep Neural Learning Classifier for Brain Tumor Detection

Abstract

Magnetic Resonance Imaging (MRI) is a scanning method which captures the anatomy and processes of human body. MRI images are significant for premature recognition of brain cancer. Thus, predicting the brain cancer disease from an MRI scan is not an easy process, because of its complexity and tumor variance. In order to address these problems, Guassian Preprocessed Projection Pursuit Regressive Mathieu Feature Extraction based Deep Neural Learning (GPPPRMFE-DNL) is introduced. GPPPRMFE-DNL Model is proposed for accurate brain tumor detection process in a short time. Gaussian smoothing filter is employed in GPPPRMFE-DNL Model to eradicate the noisy pixels from input image. Subsequently, skull stripping procedure is used for collecting brain tissue from neighbouring region. Then, the image achieved is used for dividing within the segments, to minimize the dimension of input image. Feature extraction is performed to extract the color, texture, and intensity features from the segmented region. Finally, the classification task is performed with the help of logistic activation function between the testing and training image with higher accuracy and lesser error rate. At last, the outcome is determined by an output layer. The observed results show a better analysis of GPPPRMFE-DNL, compared with the two conventional methods.