Medical images are very important in diagnosis and treatment. Medical images must be stored for monitoring of disease development. However, a confusion that may occur while storing medical images can have fatal consequences. In order to prevent confusion, watermarking of medical information such as patient, doctor, diagnosis to medical image is a better approach than storing this information in separate files. However, the structural change occurred by watermarking in the image may lead to misdiagnosis in the future. In this study, a block based watermarking method has been proposed to guarantee the authentication of medical images storing in the hospital information system. This method maps the medical image to 64 x 64 blocks and classifies these blocks with significant blocks and insignificant blocks with Gaussian function, entropy and mean squared error calculations. Thus, the watermark is embedded in the insignificant parts of the medical image and a possible misdiagnosis in the future is prevented. The method is developed by considering the high distortion caused by watermarking in grayscale images and it is applied on six different grayscale medical test images. The results compared with three variations of the classical least significant bit method and structured similarity index and peak signal to noise ratio measurements performed on medical test images.