Patient Data Hiding and Integrity Control Using Prediction-Based Watermarking for Brain MRI and CT Scan Images

Abstract

The increase of using e-Health services enables remote access, communication, and analysis of medical images and data to facilitate medical diagnostics. This has introduced potential security threats to the medical images and data. Therefore, the security of private patient information and medical images becomes an important issue that should be considered. In such a framework, reversible watermarking has been presented to enhance medical image security. Reversible watermarking is an image watermarking category, that has the capability of completely restoring the original image after the hidden data is extracted. In this paper, a reversible watermarking scheme based on adaptive prediction error is introduced, in order to ensure the confidentiality of the patient's information, verify the authenticity of the host image, and localize and recover tampered areas if the image has been modified. This scheme segments the medical image into two segments namely: Region of Interest (ROI) and Region of Non-Interest (RONI). The Patient data and the hash value of the ROI are hidden in ROI. The hash value of the RONI, tamper detection and tamper recovery data are hidden in RONI. To reduce the distortion of ROI, the most diagnostically significant region, each embeddable pixel of ROI is able to conceal one bit and each embeddable pixel of RONI could conceal two bits. From the experimental results conducted on brain MRI and CT scan images, the proposed scheme is capable to accurately locate tempered regions and recover them. Moreover, the reversibility and the high hiding capacity with very good perceptibility are proven.