Overview of the ImageCLEFmedical 2026 GANs task: Exploring fingerprint detection, privacy leakage, and synthetic medical image generation
Authors: Alexandra-Georgiana Andrei, Mihai Gabriel Constantin, Mihai Dogariu, Dzmitry Karpenka, Yuri Prokopchuk, Ahmedkhan Radzhabov, Dan-Cristian Stanciu, Liviu-Daniel Ştefan, Vassili Kovalev, Henning Müller, Bogdan Ionescu
Organizations: AI Multimedia Lab, CAMPUS Research Institute, National University of Science and Technology Politehnica Bucharest, Romania · National Academy of Science of Belarus, Minsk, Belarus · University of Applied Sciences Western Switzerland (HES-SO), Sierre, Switzerland
The 2026 ImageCLEFmedical GANs Task Exploring Fingerprint Detection, Privacy Leakage, and Synthetic Medical Image Generation continues to investigate privacy and security concerns around using patient data to generate synthetic medical images, now in its fourth edition. This edition introduces three complementary subtasks: the first extends prior editions by asking participants to detect which real images were used in training a Generative Adversarial Network to produce given synthetic outputs; the second is a novel subtask requiring participants to identify which latent vector produced each synthetic image, probing the faithfulness and reversibility of the latent-to-image mapping; the third asks participants to generate privacy-preserving CT slices, quantifying the trade-off between image fidelity and patient privacy. Ground-truth annotations and benchmark datasets of real and GAN-generated lung CT slices are provided for all subtasks. Evaluation is based on Cohen's Kappa for Subtask 1, matching accuracy for Subtask 2, and a combination of FID and Privacy Preservation Score (PPS) for Subtask 3. Overall, 37 teams registered for the task, with 10 teams submitting runs across subtasks and 6 teams submitting working notes papers. This paper presents an overview of the task setup, datasets, evaluation metrics, and summarises the approaches and results of participating teams.
Figures & tables
Team
Subtask 1
Subtask 2
Subtask 3
Affiliation
Country
Bharathi & Harish Subrahmanian and M (2026)
✓
Shiv Nadar University
India
csmorgan Shaharear et al. (2026)
✓
✓
✓
Morgan State University
USA
CVG-IBA Shaikh et al. (2026)
✓
✓
Institute of Business Administration, Karachi
Pakistan
DS@GT ARC Regina et al. (2026)
✓
Georgia Institute of Technology
USA
Agentili Gentili (2026)
✓
San Diego VA Health Care Center / UC San Diego
USA
AI Multimedia Lab* Andrei et al. (2026)
✓
✓
✓
National University of Science and Technology POLITEHNICA Bucharest
Romania
Table 1: Overview of participating teams that submitted working notes papers (* task organising team).
#
Participant
Run ID
Cohen’s κ
MCC
Acc.
Sensitivity
Specificity
F1
1
CVG-IBA
1170
-0.0243
-0.0380
0.4879
0.1037
0.8720
0.1684
2
csmorgan
467
-0.0290
-0.0613
0.4855
0.0449
0.9262
0.0802
3
Bharathi & Harish
843
-0.0383
-0.0385
0.4808
0.4290
0.5327
0.4524
4
csmorgan
468
-0.0570
-0.0612
0.4715
0.2897
0.6533
0.3541
5
csmorgan
466
-0.0626
-0.0713
0.4687
0.2299
0.7075
0.3020
6
Bharathi & Harish
857
-0.0776
-0.0779
0.4612
0.4150
0.5075
0.4351
Table 2: Results of participant submissions for Subtask 1: Detect Training Data Usage. (* task organising team)
#
Participant
Run ID
Accuracy
Correct
Wrong
1
Agentili
1008
1.0000
10000
0
2
Agentili
1009
0.9996
9996
4
3
Agentili
909
0.9916
9916
84
4
CVG-IBA
1169
0.3677
3677
6323
5
CVG-IBA
1179
0.2135
2135
7865
6
CVG-IBA
1178
0.1709
1709
8291
Table 3: Results of participant submissions for Subtask 2: Identify Corresponding Latent Vectors.(* task organising team)
#
Participant / Run
FID ↓
PPS ↑
ℓ1 (NND)
ℓ2 (MIA)
ℓ3 (Re-ID)
ℓ4 (Inv.)
1
csmorgan
3.2358
0.9086
0.0213
0.1832
0.1611
0.0000
2
DS@GT ARC – fm-unet-dp
0.3290
0.5490
0.0901
0.3797
0.9933
0.3410
3
DS@GT ARC – dp-sv-spatial-dpp
0.2639
0.4920
0.0835
0.3519
0.9732
0.6250
4
DS@GT ARC – dp-sv-geodesic-greedy
0.4272
0.4980
0.0804
0.3725
0.9866
0.5620
5
DS@GT ARC – dp-riemannian-wdpp
0.3250
0.4660
0.0880
0.3838
0.9933
0.6710
6
DS@GT ARC – dp-riemannian-skt
0.3758
0.4330
0.0874
0.3724
0.9933
0.8140
Table 4: Results of participant submissions for Subtask 3: Privacy-Preserving CT Slice Generation. (* task organising team)
Figure 1: Real training slices alongside their most similar synthetic counterparts generated by each team’s submission, as identified by the Patient Re-Identification attack (Attack 3) - part 1.
Figure 2: Real training slices alongside their most similar synthetic counterparts generated by each team’s submission, as identified by the Patient Re-Identification attack (Attack 3) - part 2.
We present a privacy-preserving framework for synthetic lung CT slice generation developed for the Image-CLEFmed GANs 2026 challenge. The approach combines Optimal Transport Conditional Flow Matching with privacy-oriented training and a post-generation "Supervisor" pipeline that filters generated candidates in learned geometric latent spaces using autoencoder embeddings, Determinantal Point Processes, and Stein Kernel Thinning. Official results show a strong realism-privacy trade-off, with the best-performing model achieving a Privacy Preservation Score of 0.549 and competitive visual fidelity with an FID of 0.3290. While the proposed geometric filtering substantially reduces nearest-neighbor memorization and membership-inference leakage, persistent patient re-identification scores indicate that preventing direct image copying is not sufficient to remove deeper patient-specific anatomical identity, highlighting an important frontier for future privacy-preserving medical image generation.
Eric Regina, Richard Arnaud, Samir Hadi Cisneros
Georgia Institute of Technology, North Ave NW, Atlanta, GA 30332
Recent advances in generative AI have made synthetic Computed Tomography (CT) images increasingly realistic, enabling promising applications in medical data augmentation while raising serious concerns about clinical safety and data trustworthiness. Detecting AI-generated CT images remains challenging for two key reasons: existing benchmarks cover only limited generation sources, and many detectors are adapted from natural-image forensics without explicitly modeling CT-specific imaging properties. In this paper, we introduce CTForensics, a dataset for detecting AI-generated CT images. CTForensics contains 75,990 2D CT images, including a dedicated test benchmark of 29,990 balanced authentic and generated samples from ten representative CT generative models spanning GAN-based and diffusion-based paradigms. We further propose the Enhanced Spatial-Frequency CT Forgery Detector (ESF-CTFD), a CT-oriented CNN framework built around a Wavelet-Enhanced Central Stem, Multi-Scale Spatial Aggregation, and a Frequency-Aware Prediction Block. The Wavelet-Enhanced Central Stem enhances local intensity correlations and high-frequency residuals, Multi-Scale Spatial Aggregation aligns anatomical features across resolutions with lightweight residual units, and the Frequency-Aware Prediction Block models global spectral artifacts. Extensive experiments on CTForensics show that ESF-CTFD achieves 96.01% mAcc and 99.96% mAP, outperforming existing methods and maintaining strong robustness under realistic perturbations with only a 0.99% average drop. Codes will be available at https://github.com/liyih/CTForensics.
Yiheng Li, Zichang Tan, Guoqing Xu +3
School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China · MAIS, Institute of Automation, Chinese Academy of Sciences, Beijing, China · Sangfor Technologies Inc., Shenzhen, China +1
Large-scale, publicly available datasets have driven advances in deep learning, but privacy and legal restrictions often limit data sharing in medical imaging. Synthetic data generation offers a privacy-friendly alternative to enable the training of high-performance models on health data. While most state-of-the-art generative models produce high-quality images, they remain computationally expensive, which limits their applicability on resource-constrained hardware. We propose StyleGANCA, the first lightweight general-purpose NCA-based generative adversarial network. The architecture integrates a StyleGAN-inspired mapping network and adaptive style modulation into a multi-scale NCA synthesis process, enabling latent-controlled image generation through iterative local interactions. We evaluate StyleGANCA on BloodMNIST and PathMNIST against adversarial, variational, diffusion, and NCA-based baselines. Experimental results demonstrate that StyleGANCA achieves competitive image quality with substantially fewer parameters than baseline architectures, achieving the best FID and KID scores on PathMNIST with only 617k parameters. Furthermore, downstream experiments show that the generated images preserve class-specific information and effectively support the training of multi-class classifiers. Our code is publicly available at: https://github.com/MECLabTUDA/StyleGANCA
Anh Thi Luu, Nick Lemke, Anirban Mukhopadhyay
Technical University of Darmstadt, Darmstadt, Germany · ImFusion GmbH, Munich, Germany