Digitisation of medical images is improving diagnosis and clinical research, but poses major challenges in terms of security and traceability. This paper proposes a hybrid solution combining digital watermarking and steganography to securely trace these images, while enabling automatic watermarking detection via adapted neural networks. We evaluated several watermarking techniques, including Discrete Cosine Transform (DCT), Discrete Wavelet Transform (DWT) and Singular Value Decomposition (SVD), in terms of their robustness to alteration, imperceptibility and impact on file size. Our results show that the DWT method applied to the LL sub-band offers a good compromise between visual quality and robustness, with almost total imperceptibility and a low impact on image size. This work provides an integrated methodology to strengthen the protection of medical data and support the development of high-performance watermarking detection algorithms in digital medical environments.