Medical data protection against illegal manipulations has become an essential and urgent issue. Unfortunately, images exchanged via networks are not absolutely protected against the preservation of integrity, authenticity and the right of use. Watermarking can play a very important role in dealing with this problem. For this reason, it becomes necessary to perform such watermarking in image regions where the disorder regarding the pixels distribution should be less when embedding secret data called watermark. In this paper, we propose a new approach of medical image watermarking in spatial domain with a non-blind way. This approach is based on one of the essential properties of image called Local Entropy (LE). The watermarking in the zones with less local entropy values guarantees a high imperceptibility of the watermark. The choice of the low local entropy is based on measuring or estimating changes in a zones or regions of the image. The watermark embedding consists only on pixels with less local entropy values since these pixels represent less disorder within the image. The results obtained are very encouraging and have been evaluated in terms of imperceptibility through the Peak Signal Noise Rate (PSNR) metric and evaluated also in terms of robustness by measuring the Correlation Coefficients (CC), Bit Error Ratio (BER) and Structural Similarity Index Measure (SSIM) between the original watermark and the extracted ones.