With the growth of globalisation, security has been a major concern in digital exchange of medical information. Data hiding, also referred as watermarking, was introduced to authenticate medical data which is sent over the network. Watermarking of electronic patient reports (EPR) for telemedicine demands robust, imperceptible, high payload techniques. Blood vessel extraction is mainly used for diabetic retinopathy (DR) for automatic extraction and classification of severity of diseases. In our current work, blood vessel from fundus images is extracted using K-means segmentation. Afterwards, EPR is hidden using interpolation and trigonometric functions in fundus image. Finally, blood vessel is extracted from the watermarked image to measure the changes in accuracy of the proposed system. The percentage difference (< 0.25%) of accuracy in fundus images before watermarking and after watermarking claims the retention of devalorisation of vessel extraction accuracy measurement. High peak signal to noise ratio (PSNR) value (> 36) and high correlation (> 87%) between original and watermarked retinal image, establishes the robustness of the proposed non-blind watermarking method.