A hybrid local prediction-error based difference expansion reversible watermarking algorithm for hiding data into medical images is presented.The performance of hybrid local algorithm with median, median edge, gradient adjusted and local prediction algorithms is compared.The hybrid local prediction algorithm has the highest frequency of zero prediction error.The PSNR and embedding capacity are improved in the hybrid local algorithm. Embedding secret information into a cover media and extracting the information and the image without any distortion is known as reversible watermarking (RW). This paper analyzes the performance of hybrid local prediction error-based RW using difference expansion (DE). The cover medical image is split into non-overlapping blocks. The border pixels in each block are predicted using median edge detection (MED) prediction. The other pixels in the block are predicted using least square prediction, and the prediction error is expanded. The secret data are embedded into the cover medical image corresponding to the prediction error using the DE method. The predictor is also embedded into the cover medical image to recover the data at detection without any additional information. The simulation results show that, this method achieves better watermarked image quality and high embedding capacity when compared to other classical prediction methods: Median, MED, Rhombus and Gradient Adjusted Prediction. Display Omitted