HexMIL: Hierarchical Attention MIL for Ante-Hoc Explainable Detection of AI-Manipulated CT Volumes
2026-08-05 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionMultimedia
AI summaryⓘ
The authors developed HexMIL, a new tool to detect medical deepfakes—fake medical images created by AI—that works better than existing methods. HexMIL breaks down 3D medical scans into parts and uses special attention modules to spot fake regions without needing detailed pixel-level labels. It also directly shows which areas influenced its decision, making it easier to understand why it flagged an image. Tested on medical datasets, HexMIL performed better at identifying fake images from unknown AI models and was more accurate at locating the fake parts than previous methods.
medical deepfakesdeep generative modelsCT volumemultiple instance learninggated attention3D attention volumecross-generator generalizationAUCF1 scorelocalization metrics
Authors
Orazio Pontorno, Luca Guarnera, Zahid Akhtar, Sebastiano Battiato
Abstract
The emergence of medical deepfakes, i.e., medical images manipulated by deep generative models, poses a significant threat to clinical workflows. However, existing detectors suffer from two critical limitations: poor generalization to unseen generative architectures for manipulation detection and lack of interpretability. In this context, we present HexMIL (Hierarchical EXplainable Multiple Instance Learning), a mask-free medical deepfake detector that simultaneously addresses both limitations using only binary volume-level supervision. HexMIL decomposes each CT volume into a two-level hierarchy of patches and slices, aggregated via independent Gated Attention modules whose weights are directly combined into a full-resolution 3D attention volume that localizes the manipulated sub-region without any pixel-level annotation. Unlike post-hoc methods such as Grad-CAM, HexMIL's attention weights constitute the exact forward computation driving the classification decision, providing ante-hoc and structurally faithful spatial attribution. We evaluate HexMIL on M3DSynth and CT-GAN datasets under a rigorous cross-generator generalization protocol, training on a single generative architecture and testing on unseen ones. HexMIL outperforms all baselines by $+9.1$ AUC and $+9.4$ F1 in out-of-domain classification, and achieves the best average IoU and Pointing Game score in localization. Project page: opontorno.github.io/hexmil.