Robust Medical Image Watermarking and Brain TumorSegmentation Using Multi-Domain Transforms And Deep ‎Learning

This observation introduces a unified framework that guarantees both the stable watermarking of medical pics and ‎effective brain tumor segmentation. By leveraging a hybrid approach that integrates deep learning with a ‎couple of area transformation strategies, the device complements photograph security and diagnostic precision. ‎The approach embeds MRI scans into host photographs through a fusion of Discrete Wavelet Transform, Discrete ‎Cosine Transform, and Singular Value Decomposition. The embedding strength is adaptively optimized using ‎Black Widow Optimization guided by a Genetic Algorithm, enhancing robustness and imperceptibility. DWT ‎decomposes the images, DCT shifts them to the frequency domain, and SVD modifies singular values with an ‎adaptive power factor. Inverse operations reconstruct the watermarked image, authenticated using a unique ‎code via the ThingSpeak IoT platform. Successful authentication enables watermark extraction through ‎inverse SVD, DCT, and DWT, yielding high fidelity (PSNR > 34 dB, SSIM > 0.91). A Hybrid U-Net-based deep ‎mastering segmentation method is employed to extract mind tumors from watermarked clinical snapshots for ‎diagnostic analysis. Morphological operations and boundary extraction refine the segmented regions. Tumor ‎stages—initial, middle, or advanced—are determined based on pixel count. This dual-purpose framework ‎ensures secure embedding and reliable transmission of sensitive medical data, while also providing accurate and ‎efficient brain tumor detection. The system is highly relevant for applications in telemedicine, secure medical ‎data sharing, and remote diagnosis‎.