3D Medical Images Segmentation and Securing Based GAN Architecture and Watermarking Algorithm Using Schur Decomposition

Recently, due to the exponentially evolution of medical imaging system, 3D medical image segmentation be a useful tool for cancer detection and localization. In fact, several convolution neural networks with different architectures are used for this purpose. Those techniques significantly improved the efficiency and the performance of automatic 3D medical image segmentation with high computation and memory requirements. Meanwhile, since the performance using a computer system, the privacy and the security of patient data are vulnerable to cyber-attacks. For this, DICOM data must be a primary concern in medical applications among hospitals and clinics. In this work, we proposed a hybrid technique for 3D image segmentation and protection. In fact, this approach we propose two steps. The first one is a 3D segmentation operation, which performed with the GAN neural network architecture, and the second one is a patient data protection technique using an appropriate watermarking algorithm. The simulation results on real application prove the efficiency of this method. In fact, the obtained findings are interesting that combine the method of deep learning the GAN in segmentation of medical images and securing them with the appropriate algorithm.