The K-Space Signature: Frequency-Domain Representation Learning for Medical Deepfake Detection

2026-07-31

AI Summary

This study addresses the growing threat of medical deepfakes to public health by proposing the K-Space Signature (KSS) method, which jointly models hardware- and generation-induced artifacts in the frequency domain. By introducing a global anatomical prior in the log power spectral density space, the approach effectively mitigates variations due to anatomical differences. A spatially unbiased 3D MLP-Mixer architecture, integrated with ArcFace metric learning, enables highly robust forgery detection. This work pioneers the fusion of frequency-domain representations with anatomical priors for medical deepfake identification and achieves zero-shot generalization across imaging devices. Evaluated on multi-center 3D MRI data, the model attains detection accuracy and ROC-AUC exceeding 0.99, with zero-shot accuracy of 0.93 on unseen scanners.

Abstract

In medical imaging, generative models are increasingly deployed to synthesize realistic data and augment limited datasets. Unfortunately, while beneficial for privacy-preserving data sharing, these synthesized images can be repurposed for malicious intents, threatening public health through the creation of Medical Deepfakes. To address this threat, we introduce the K-Space Signature (KSS), a novel forensic framework that isolates hardware and generative traces within the spectral domain. By shifting analysis to the frequency domain, the KSS suppresses macroscopic anatomical variance by subtracting an empirical global anatomical prior computed in the Logarithmic Power Spectral Density (Log-PSD) space. To effectively process these globally distributed spectral artifacts without the local spatial bias inherent to Convolutional Neural Networks, we pair the KSS representation with a novel 3D MLP-Mixer architecture equipped with an ArcFace metric-learning head. Extensive experiments on multi-center 3D MRI datasets demonstrate that this combined approach achieves exceptional detection performance, exceeding 0.99 Accuracy and ROC-AUC on multi-generator synthetic datasets. Furthermore, the framework exhibits robust zero-shot generalization, maintaining strong discriminative power (up to 0.93 Accuracy) on independent datasets acquired from entirely unseen scanners. To ensure full reproducibility, the complete source code and pre-trained models will be made publicly available upon acceptance.

Problem

• Medical Deepfake
• Deepfake Detection
• Generative Models
• Medical Imaging
• Synthetic Data

Innovation

• K-Space Signature
• frequency-domain representation
• medical deepfake detection
• 3D MLP-Mixer
• zero-shot generalization