A New Deep Learning Architecture for Brain Tumor Segmentation Using <scp>MRI</scp> Images Based on Blockchain

ABSTRACT An automatic segmentation of brain tumors from medical images is significant for diagnosis and treatment planning. Magnetic resonance imaging (MRI) is widely used by surgeons, radiologists, and medical specialists for brain segmentation. To enhance the reliability and security of the segmentation process, this study introduces a blockchain‐based Recursive Residual Blocks and Pyramid Pooling Segmentation Network (RP‐Segnet) for brain tumor segmentation using MRI. In the proposed framework, blockchain technology is employed to ensure the integrity and traceability of the parameters of the model by encrypting and verifying layer‐specific data, such as weights and biases. This prevents unauthorized tampering and enables secure handling of model data, thereby supporting robust and trustworthy segmentation. Initially, input MRI is acquired from the dataset and is transferred to the preprocessing step, wherein the unwanted images are eliminated by employing Gaussian filtering. Once the preprocessing is completed, the brain image is transferred to the brain tumor segmentation module, which separates the tumor region from the background of the tumor image, which is performed using the proposed RP‐Segnet. The RP‐Segnet is designed by the formation of Recursive residual blocks and a Pyramid pooling Network (RP‐Net) and Segmentation Network (SegNet). Moreover, the efficacy of the proposed RP‐Segnet for brain tumor segmentation is determined based on evaluation measures, such as segmentation accuracy, dice coefficient, and Jaccard similarity, and it has attained maximal values of 92.04%, 95.07%, and 95.03%.