Protecting medical images in medical applications is considered a delicate task and needs to be carefully managed. Generally, the demand is to protect the data of the patient without changing the content of the actual diagnostic data. Zero-watermarking (ZWM) technique provides an elegant solution as it logically links copyright information to essential image properties instead of hiding it in the image pixels themselves. However, there is a trade-off in performance; this is because the existing methods fail to handle image rotation attacks. For example, a small tilt of the image can render the watermark ineffective due to a drop in the feature extraction process. In this paper, the proposed method tackles the aforementioned challenge using Dual-Tree Complex Wavelet Transform (DT-CWT) to generate stable, shift-invariant features from the low-frequency components. Then, the extracted features are processed with Improved Differential Entropy (IDE) to resist common attacks. In addition, the most important part is the blind geometric correction system, where it automatically detects and corrects rotation or reflection by analyzing statistical moments and image distortion, which is performed without the need for the comparison of original image. Finally, the security is enhanced using Arnold scrambling and logistic mapping before mapping the watermark to the extracted features. The developed method is resilient to noise, filtering, JPEG compression, and crucially, geometric attacks. The results show that the Normalized Correlation (NC) scores above 0.99 under different attacks including heavy rotation, solving a long-standing vulnerability in ZWM research. For medical image protection, this developed method is considered reliable, secure, and practical for telemedicine applications.