An improved framework for brain tumor analysis using MRI based on YOLOv2 and convolutional neural network

Abstract Brain tumor is a group of anomalous cells. The brain enclosed in more rigid skull. abnormal cell grows and initiates tumor. Detection complicated task due to irregular shape. proposed technique contains four phases, which are lesion enhancement, feature extraction selection for classification, localization, segmentation. magnetic resonance imaging (MRI) images noisy certain factors, such as image acquisition, fluctuation field coil. Therefore, homomorphic wavelet filer used noise reduction. Later, extracted features from inceptionv3 pre-trained model informative selected using non-dominated sorted genetic algorithm (NSGA). optimized forwarded classification after slices passed YOLOv2-inceptionv3 designed the localization region that depth-concatenation (mixed-4) layer supplied YOLOv2. localized McCulloch's Kapur entropy method segment actual region. Finally, validated on three benchmark databases BRATS 2018, 2019, 2020 detection. achieved greater than 0.90 prediction scores segmentation lesions. Moreover, outcomes superior compared existing methods.