AMIF: Authorizable Medical Image Fusion Model with Built-in Authentication

Abstract

Multimodal image fusion enables precise lesion localization and characterization for accurate diagnosis, thereby strengthening clinical decision-making and driving its growing prominence in medical imaging research. A powerful multimodal image fusion model relies on high-quality, clinically representative multimodal training data and a rigorously engineered model architecture. Therefore, the development of such professional radiomics models represents a collaborative achievement grounded in standardized acquisition, clinical-specific expertise, and algorithmic design proficiency, which necessitates protection of associated intellectual property rights. However, current multimodal image fusion models generate fused outputs without built-in mechanisms to safeguard intellectual property rights, inadvertently exposing proprietary model knowledge and sensitive training data through inference leakage. For example, malicious users can exploit fusion outputs and model distillation or other inference-based reverse engineering techniques to approximate the fusion performance of proprietary models. To address this issue, we propose AMIF, the first Authorizable Medical Image Fusion model with built-in authentication, which integrates authorization access control into the image fusion objective. For unauthorized usage, AMIF embeds explicit and visible copyright identifiers into fusion results. In contrast, high-quality fusion results are accessible upon successful key-based authentication.

Figures (9)

• Figure 1: Why AMIF. Medical images are privacy-sensitive and should not be freely distributed. However, existing fusion models typically ignore data leakage risks and model intellectual property protection. AMIF is trained on standardized, clinically representative multimodal data to learn fusion with built-in copyright protection. It outputs a watermarked fused image by default and restores a watermark-free result only with a valid key.
• Figure 2: AMIF enables authorizable multimodal medical image fusion with built-in authentication. A shared Restormer encoder extracts common features and private encoders capture modality-specific cues, which are fused into a multimodal representation. CCWM first generates an internal content-aware watermark that carries implicit content cues. C-SAMIC then injects it into the protected features via tightly-coupled wavelet-domain invertible coupling, producing a copyrighted output and an associated key. With a valid key, the inverse mapping removes the watermark and recovers a high-quality watermark-free fusion result for authorized use. The whole pipeline works in a unified and cooperative manner.
• Figure 3: Visualization of fusion results in the unauthorized mode. AMIF outputs fused images with clear visible watermarks for copyright protection.
• Figure 4: Qualitative results of authorized (watermark-free) fusion images produced by AMIF compared with different fusion models on three multimodal datasets. A and B denote the two input modalities.
• Figure 5: Visual results under watermark removal attacks. Columns A and B show the two input modalities, followed by the copyrighted fused image and the attacked outputs of SLBR, WDNET, and PATCHWIPER. Forced watermark removal in the unauthorized setting introduces distortions and artifacts, degrading the fused content.