An Efficient Brain Tumor Prediction Using Pteropus Unicinctus Optimization on Deep Neural Network

Abstract

Human brain tumors are now the most serious and horrible diseases for people, causing certain deaths. The patient’s life also becomes more complicated over time as a result of the brain tumor. Thus, it is essential to find tumors early to safeguard and extend the patient’s life. Hence, new improvements are highly essential in the techniques of brain tumor detection in medical areas. To address this, research has introduced automatic brain tumor prediction using Pteropus unicinctus optimization on deep neural networks (PUO-deep NNs). Initially, the data are gathered from the BraTS MICCAI brain tumor dataset and preprocessing and ROI extraction are performed to remove the noise from the data. Then the extracted RoI is forwarded to the fuzzy c-means (FCM) clustering to segment the brain image. The parameters of the FCM tune the PUO algorithm so the image is segmented into the tumor region and the non-tumor region. Then the feature extraction takes place on ResNet. Finally, the deep NN classifier successfully predicted the brain tumor by utilizing the PUO method, which improved the classifier performance and produced extremely accurate results. For dataset 1, the PUO-deep NN achieved values of 87.69% accuracy, 93.81% sensitivity, and 99.01% specificity. The suggested PUO-deep NN also attained the values for dataset 2 of 98.49%, 98.55%, and 95.60%, which is significantly more effective than the current approaches.