SBTC-Net: Secured Brain Tumor Segmentation and Classification Using Black Widow With Genetic Optimization in IoMT

People around the globe are suffering from different types of brain tumors. So, early prediction of brain tumors can save human lives. This work focused on implementing a secured brain tumor classification network (SBTC-Net) using transfer learning methods. Initially, security is achieved by performing the medical image watermarking (MIW) operation using translation invariant wavelet transform (TIWT). Here, the watermarking process covers a patient's source MRI brain tumor image with an unknown medical image (cover image). Then, this watermarked image is transmitted over the Internet of Medical Things (IoMT) environment. Here, the attackers are unable to visualize the source image. So, the source brain tumor image is transmitted over a secured environment. At the receiver of IoMT, the segmentation operation is performed using the transfer learning-based Recurrent U-Net (RU-Net) model, which localizes the exact area of the tumor. In addition, multilevel features are extracted using the black widow optimization-genetic algorithm (BWO-GA), which selects the best features using naturally inspired properties. Further, transfer learning based AlexNet trains with the optimal features, classifying benign and malignant tumor tumors. Finally, the simulation results show that the proposed SBTC-Net resulted in superior watermarking, segmentation, and classification performance in terms of subjective visualization and objective metrics compared to state-of-the-art approaches. The proposed SBTC-Net achieved 99.97% of segmentation accuracy and 99.98% of classification accuracy on the BraTS-2020 dataset.